Authors - Sheela S. J, Rajeshwari B. S, Harsha M, Subhash T. D, Tejas H. S, Thanmaya Ganesh C. S, Harsha S. M, Keerthana T. V Abstract - One of the leading causes of death Globally cardiovascular diseases (CVDs). 2019 key Statistics on CVDs is as follows: Total Deaths: 17.9 million people, 32% of global deaths, 85% of CVD deaths which approximately 15.2 million deaths from heart attacks and strokes. Hence, early diagnosis plays a crucial role in reduction of heart related diseases. Usually, the healthcare professionals collect the initial cardiac data using their quintessential instrument called stethoscope. Traditionally, these stethoscopes have significant drawbacks such as weak sound enhancement and limited noise filtering capabilities. Moreover, the low frequency signals such as below 50 Hz may not be heard because of the variation in sensitivity of a human ear. Hence, the usage of conventional stethoscopes requires experienced medical practitioners. In order to overcome these limitations, it is necessary to develop a device which is more sophisticated than conventional stethoscopes. In this context, the proposed work aims in the development of digital stethoscope which has the capability of displaying heart and lungs sound separately. Further, the proposed digital stethoscope permits to document, convert and transmit heart and lungs sounds to dB range digitally thereby reducing unnecessary travelling to medical facilities. The proposed stethoscope results are compared and validated with conventional techniques.
Authors - Prem Gaikwad, Parth Masal, Mandar Kulkarni, Mousami P. Turuk Abstract - Visual Language Models (VLMs) are an emerging technology that integrates computer vision with natural language processing, offering transformative potential for healthcare. VLMs significantly enhance disease detection, diagnosis, and report generation by enabling automated analysis and interpretation of medical images. These models are designed to support healthcare professionals by streamlining workflows, improving diagnostic accuracy, and assisting in clinical decision-making. Applications include early disease detection through image analysis, automated report generation, and integration with electronic health records (EHR) for personalized medicine. Despite their promise, challenges such as data privacy, interpretability, and the scarcity of labelled datasets remain. However, ongoing advancements in AI-driven medical systems and the integration of multimodal data can potentially revolutionize patient care and operational efficiency in healthcare settings. Addressing these challenges is crucial for realizing the full potential of VLMs in clinical practice.
Authors - Kamini Solanki, Nilay Vaidya, Jaimin Undavia, Jay Panchal Abstract - Polycystic ovary disease (PCOD) is a condition in which the ovaries of women of childbearing age produce too many immature or partially mature eggs. As time passes, these eggs develop into cysts within the ovaries. These cysts can lead to enlargement of the ovaries and an elevated production of male hormones (androgens). Consequently, this hormonal imbalance can result in a range of issues like fertility challenges, irregular menstrual cycles, unanticipated weight gain, and various other health complications. The associated symptoms often exert a long-term impact on both the physical and mental well- being of affected women. Statistics indicate that approximately 34% of individuals facing PCOD also grapple with depressive symptoms, while almost 45% experience anxiety. The primary object of this proposed framework is to detect and classify PCOD disease from standard X-ray pictures with assistance of volume datasets using deep learning model. Polycystic Ovary Disease (PCOD) significantly affects women's reproductive health, leading to various long-term complications. This work introduces a novel framework for automated PCOD detection using integrating Convolutional Neural Networks (CNN) with deep learning, applied to ultrasound imaging. Unlike traditional diagnostic methods, which rely on manual interpretation and are prone to subjectivity, the proposed system leverages the powerful feature extraction capabilities of CNNs to classify infected and non-infected ovaries with 100% accuracy. This high level of precision outperforms existing models and can be seamlessly integrated into clinical workflows for real-time diagnosis during sonography, facilitating early detection and improved fertility management. By focusing on a deep learning approach, this work provides a scalable, reliable, and automated solution for PCOD diagnosis, marking a significant advancement in the use of medical imaging with artificial intelligence.
Authors - Poornima E. Gundgurti, Shrinivasrao B. Kulkarni Abstract - Latent fingerprints play a crucial role in forensic investigations, driven by both public demand and advancements in biometrics research. Despite substantial efforts in developing algorithms for latent fingerprint matching systems, numerous challenges persist. This study introduces a novel approach to latent fingerprint matching, addressing these limitations through hybrid optimization techniques. Recognizing latent fingerprints as pivotal evidence in law enforcement, our comprehensive method encompasses fingerprint pre-processing, feature extraction, and matching stages. The proposed latent fingerprint matching utilizes a novel approach named as, Randomization Gravity Search Forest algorithm (RGSFA). Acknowledging the shortcomings of traditional techniques, our method enhances convergence speed and performance evaluation by integrating weighted factors. Precision, recall, F-measure, and recognition rate serve as performance metrics. The proposed approach has a high recognition rate of 99.9% and is successful in identifying and matching latent fingerprints, furthering the development of biometric-based personal verification techniques in forensic science and law enforcement. Experimental analyses, using publicly accessible low-quality latent fingerprints from FVC-2004 datasets, demonstrate that the proposed framework outperforms various state-of-the-art approaches.
Authors - Krunal Maheriya, Mrugendra Rahevar, Martin Parmar, Deep Kothadiya, Arpita Shah Abstract - Plant diseases pose a significant threat to agricultural output, causing food insecurity and economic losses. Early detection is crucial for effective treatment and control. Traditional diagnosis methods are labor intensive, time-consuming, and require specialized knowledge, making them unsuitable for large scale use. This study presents a novel approach for classifying cassava leaf diseases using stacked convolutional neural networks (CNNs). The proposed model leverages pre-trained ResNet-18 features to enhance feature learning and classification accuracy. The dataset includes images of cassava leaves with various diseases, such as Cassava Mosaic Disease (CMD), Cassava Green Mottle (CGM), Cassava Bacterial Blight (CBB), and Cassava Brown Streak Disease (CBSD). Our method begins with data preparation, including image augmentation to increase robustness and variability. The ResNet-18 model is then used to extract high-level features, which are then fed into a stacked CNN architecture made up of pooling layers, several convolutional layers, and non-linear activation functions. A fully connected layer is then used for classification. Experimental results demonstrate high accuracy in categorizing cassava leaf diseases. The proprietary stacked CNN architecture combined with pre-trained ResNet-18 features offers a significant improvement over conventional machine learning and image processing methods. This study advances precision agriculture by providing a scalable and effective method for early disease identification, enabling farmers to control diseases more accurately and promptly, thereby increasing crop yield. The findings point to the promise of deep learning techniques in agricultural applications and provide directions for further study to create more complex models for the classification and diagnosis of plant diseases.
Authors - Ruchi Tripathi, Anjan Mishra, Subrata Mondal, Arunangshu Giri, Dipanwita Chakrabarty, Wendrila Biswas Abstract - Agricultural product shares a significant part of retail industry. The growing popularity of digital ecosystem can immensely affect agricultural sector as well. The consumers and retailers both can get benefited from Internet of Things or IoT, as it has a vast application in agricultural product retailing. IoT helps a retailer to establish an efficient supply chain with minimum wastage without compromising with quality. On the other side, it delivers authentic real-time information to the consumers, so that they can take efficient decision. This study has identified some factors that yields decision satisfaction to the consumers through application of IoT in agricultural product retailing sector.
Authors - Khush Zambare, Amol Wagh, Sukhada Mahale, Mayank Sohani Abstract - In the all-digital world of today, the search engines are more of entry points to knowing most things. In this regard, most search engines often service the general user; most other needs, specific to a profession, go unattended. Use of "Amazon" will return results for the e-commerce giant, even when the user is the environmental scientist looking for something about the Amazon rainforest or the cloud developer searching for Amazon Web Services (AWS). This generic approach leads to inefficiencies as users need to sift through lots of useless information. This paper allows for a browser extension that personalizes the result of search on Artificial Intelligence and Machine Learning, with the aim of catering to individual users, based on their profession, interests, and specific needs. The solution dynamically re-ranks the search results as it learns from user behavior and search patterns to provide the most relevant information to save the precious time of the users. The paper will discuss current trends in SEO, AIML applications, and personalization techniques to outline how this solution can revolutionize the search engine experience.
Authors - Suruchi Pandey, Hemlata Vivek Gaikwad Abstract - The rapid shift in the integration of AI in various sector for more personalized and efficient training. This research explores into the potential of AI in various training methods, the challenges and vast opportunities of learning and growth while using it. The potential for AI-driven training is vast, spanning fields like corporate, healthcare, education, and the military. This study examines how emerging technologies like virtual reality, augmented reality, and simulation-based training can personalize learning experiences, enhance skill development, and provide real-time feedback. It also addresses critical challenges to implementing AI in training, such as costs and data privacy concerns. Additionally, the paper discusses how AI-enabled training could transform traditional learning and development practices, opening up new possibilities for advanced, adaptive learning methods.
Authors - Shripad Kanakdande, Atharva Kanherkar, Ayush Dhoot, P.B.Tathe Abstract - Efficient inventory forecasting and waste management are essential for streamlining supply chains and cutting expenses, particularly in sectors like retail and food services where inadequate stock management can lead to large losses and environmental damage. This study presents a data-driven approach to inventory prediction that makes use of sophisticated machine learning models that evaluate past data, sales patterns, and seasonal fluctuations. The model seeks to increase demand forecasting accuracy by utilizing predictive skills, which would ultimately result in improved stock management and customer satisfaction. In order to help organizations reduce waste and increase resource efficiency, it also focuses on improving waste management through real-time monitoring and forecasting of surplus inventory. Furthermore, combining sustainable practices with predictive analytics promotes long-term corporate viability while minimizing environmental harm. In addition to increasing operational effectiveness, this all-encompassing strategy supports more general environmental sustainability goals. The suggested framework gives businesses a practical way to optimize and streamline their supply chain operations while fulfilling sustainability goals by offering a complete solution that can minimize the ecological footprint and the costs associated with keeping inventory.
Authors - U.Sakthi, Aman Parasher, Akash Varma Datla Abstract - This work seeks to classify various ship categories on the high-resolution optical remote sensing dataset known as FGSC-23 using deep learning models. The dataset contains 23 types of ships, but for this study, six categories are selected: Medical Ship, Hovercraft, Submarine, Fishing Boat, Passenger Ship and Liquified Gas Ship. The adopted ship categories were thereafter used to train four deep learning models which included VGG16, EfficientNet, ResNet50v2, and MobileNetv2. The accuracy, precision, and AUC parameters were used to evaluate the models where the best one, the ResNet50v2, was set up as accurate. Using these models, it should be possible to achieve a practical deployment aiming at fine-grained classification of ships that will contribute to enhancing maritime surveillance techniques. ResNet50v2 model had the highest precision of 0.9058 and on the other hand MobileNetv2 had the highest AUC of 0.9932. The analysis of the identified models is performed further in this work to illustrate their advantages and shortcomings in adherence to fine-grained ship classification tasks. Based on this research, the practical implications transcend theoretical comparisons of performance metrics, as useful information is provided to improve security applications in the maritime domain, surveillance, and monitoring systems. Categorization and identification of ships is a very important process in going global maritime business because it is used in decision-making processes in fields like security and surveillance, fishing control, search and rescue and conservation of the environment. The models highlighted are namely ResNet50v2 as well as MobileNetv2, proved to be robust in real-time applications such scenarios because of their ability to accurately and proficiently distinguish the differences between the ship types. In addition, this study suggests the luminal possibility of doing further improvement on these models using data enhancement strategies like transfer learning, data augmentation, and hyperparameter optimization which would enable it to perform impressively on any other maritime datasets. Therefore, the outcomes are beneficial for furthering work in automated ship detection and classification and is important toward enhancing the overall effectiveness and safety of navies across the globe.
Authors - Sandeep Shinde, Parth Kedari, Atharva Khaire, Shaunak Karvir, Omkar Kumbhar Abstract - With the use of cutting-edge technologies like Flask, web technology, API rendering, Make It Talk, and machine learning (ML), an AI smart tutor bot is being implemented with the goal of giving users an engaging and customized learning experience. The bot uses machine learning techniques to analyze responses and generates quiz-style questions with multiple-choice possibilities and extended answers. This allows for quick feedback. Additionally, it has an interview mode in which the user engages with an AI avatar that reads their body language and facial emotions. Using written material and specialized alphabets, the AI avatar is dynamically educated, gaining comprehensive knowledge and an accurate evaluation of user performance. The research article goes into detail about the system architecture, how different technologies were integrated, and the process for training the avatar and gauging user response. Through user feedback and experimental trials, the AI Smart Tutor Bot's performance is assessed, showcasing its potential as an advanced teaching tool that can adapt to each student's unique learning needs while boosting comprehension and engagement.
Authors - Bhagyashree D. Lambture, Madhavi A. Pradhan Abstract - Phytochemical qualities, geographic information, environmental conditions, and traditional medicinal knowledge are some of the sources of information that are incorporated into this research project, which presents a comparative examination of machine learning (ML) algorithms for the qualitative evaluation of medicinal plants. In order to categorize and forecast the medicinal value of plants based on multi-modal data, the purpose of this study is to investigate the effectiveness of various machine learning algorithms. For the purpose of determining which method is the most effective for evaluating complicated and diverse datasets, a full evaluation is carried out utilizing well-known machine learning models. These models include decision trees, random forests, support vector machines, and deep learning algorithms. Key criteria including as accuracy, precision, recall, F1-score, and computing efficiency are utilized in order to evaluate the levels of performance achieved by each method. For the purpose of gaining a deeper comprehension of the role that each data source plays in determining the medicinal potential of plants, the value of features and their interpretability are also investigated. A basis for the ongoing development of AI-driven tools in pharmacological research and plant-based drug discovery is provided by the findings of this comparative analysis, which offer vital insights into the usefulness of machine learning for medicinal plant assessment. Contributing to the expanding fields of computational botany and natural product science, the purpose of this study is to improve the precision and effectiveness of the evaluation of medicinal plants.
Authors - Azhar Abbas, Farha Abstract - Fraudulent claims in the insurance industry lead, to significant financial losses and negatively affect both policyholders and insurance firms. Machine learning has proven to be revolutionizing fraud detection since it is more than just performing the ordinary rule-based systems while automating and optimizing detection processes. The current work proposes a novel hybrid approach that combines supervised and unsupervised techniques in machine learning with applications in accurately and robustly detecting insurance fraud. Three primary models include in the framework are Decision Tree, Random Forest, and Voting Classifier, which improve detection performance on real-world datasets. In addition, an embedding-based model interprets sequential claims data, and a statistically validated network is used to detect patterns of collusion and fraud among related entities. Extensive experimentation was conducted using large-scale motor, and general insurance datasets and showing that the proposed hybrid model achieved an accuracy of 89.60%. Hyperparameter tuning and data preprocessing were used to further refine the model's performance; it was able to counterbalance all issues brought forth by imbalances, complexities, and complexities due to variations in fraud types. The methodology outperformed the existing models, better at identifying rare sophisticated cases of fraud. The practical implications of deploying machine learning models in the insurance sector are also discussed from the angle of best practices for data governance, model interpretability, and stakeholder trust. In Future this work will be improved by incorporating real-time analytics to provide quicker detection, enhancing interpretability features, and adapting the model to emerging fraud patterns in evolving data environments.
Authors - Poonam Yadav, Meenu Vijarania, Meenakshi Malik, Neha Chhabra, Ganesh Kumar Dixit Abstract - Parkinson's disease is aging-associated degenerative brain illness that results in the degeneration of certain brain regions. Early medical diagnosis of Parkinson's disease (PD) is difficult for medical professionals to make with precision. Magnetic Resonance Imaging (MRI) and single-photon emission computed tomography, or SPECT are two medical imaging strategies that can be used to non-invasively and safely assess quantitative aspects of brain health. Strong machine learning and deep learning methods, along with the efficiency of medical imaging techniques for evaluating neurological wellness, are necessary to accurate the identification of Parkinson's disease (PD). In this study, we have used dataset of MRI images. This study suggests three deep learning models: ResNet 50, MobileNetV2 and InceptionV3 for early diagnosis of PD utilizing MRI database. From these three models, MobileNetV2 demonstrated superior accuracy in training, testing and validation with a rate of 99%, 94%, and 96%, respectively. With its effectiveness and precision, MobileNet V2 has a lot of potential for PD identification using MRI scans in the future. We may further advance the development of dependable and easily accessible AI-powered solutions for early diagnosis and better patient care by tackling the issues and investigating the above-mentioned future paths.
Authors - Aye Thiri Nyunt, Nishi Vora, Devanshi Vaghela, Brij Kotak, Ravi Chauhan, Kirtirajsinh Zala Abstract - This paper is an AI and Machine Learning Algorithm - based dualistic Gesture-to-Speech and Speech-to-Gesture framework. The core of this initiative is to enable machines and humans to converse with each other by enabling the translation of physical body movements into reasonable speech and vice versa. We used deep learning models- Convolutional Neural Networks (CNN)- to train our system using a dataset consisting of human gestural movements and the relevant speech patterns. For the Gesture-to-Speech module, real-time gesture recognition and interpretation were used, which involved computer vision and were implemented to interpret gestures into speech output containing words and phrases representing the message illustrated by the gestures. The Speech-to-Gesture module, on the other hand, uses speech as input to produce context-related gestures-the main purpose of which is to improve user interaction and experiences. In the system, multiple applications were tested, including sign language and webcams. Further research will try to extend the flexibility of the system to include various languages, cultural backgrounds and characteristics of individual gesture styles which eventually has a high level of customization. We had designed the CNN architecture for real-time gesture recognition and taken care of data preprocessing as well to increase accuracy concerning different types of gestures. We created Gesture-to-Speech translation with the use of an LSTM, then added in a Text-to-Speech engine for it to have a very natural sound. We then worked on Speech-to-Gesture and even refined the gestures through a CNN-based network, to ensure transitions are very fluid. Everything was coordinated such that there would be synchronous gestures and speech for extremely natural real-time interaction. We coached on how one would integrate, test, and further optimize models with dropout and batch normalization for higher performance.
Authors - Varun Maniappan, Praghaadeesh R, Bharathi Mohan G, Prasanna Kumar R Abstract - This paper constitutes a comprehensive review of how language models have changed, focusing specifically on the trends toward smaller and more efficient models rather than large, resource-hungry ones. We discuss technological progress in the direction of language models applied to attention mechanisms, positional embeddings, and architectural enhancements. The bottleneck of LLMs has been their high computational requirements, and this has kept them from becoming more widely used tools. In this paper, we outline how some very recent innovations, notably Flash Attention and small language models (SLMs), addressed these limitations by paying special attention to the Mamba architecture that uses state-space models. Moreover, we describe the emerging trend of open-source language models, reviewing major technology companies efforts such as Microsoft’s Phi and Google’s Gemma series. We trace here the evolution from early models of transformers to the current open-source implementations and report on future work to be done in making AI more accessible and efficient. Our analysis shows how such advances democratize AI technology while maintaining high performance standards.
Authors - S.K. Manjula Shree, Shreya Vytla, J. Santharoopan, Harisudha Kuresan, A.Anilet Bala, D.Vijayalakshmi Abstract - The goal is to use a Random Forest classifier to categorize future price movements as "up" or "down" in order to forecast stock market trends. In order to guide investing strategies, this model will examine pertinent attributes and previous stock data. The effectiveness of Logistic Regression, Support Vector Machines (SVM), and Random Forest Classifier in forecasting stock market movements is compared in this study. The ensemble approach Random Forest is very resilient under erratic market situations since it is excellent at handling noisy, complex data and capturing non-linear patterns. SVM performs best on smaller, more structured datasets, however noise and non-linearity might be problematic. Despite its simplicity and interpretability, logistic regression is constrained by its linear character and finds it difficult to account for the dynamic, non-linear behavior of stock prices. In recall focused tasks, logistic regression is helpful because it performs well in identifying true positives (such preventing missed opportunities in stock predictions). SVM's reliance on kernel functions makes it computationally expensive, but it can also be helpful when handling smaller datasets with clear patterns and where accuracy is needed. All things considered, Random Forest offers the greatest results around 99% especially for difficult stock market prediction assignments.
Authors - Umakant Singh, Ankur Khare Abstract - This paper aims to find the optimal model for stock price forecast. In examining the different approaches and aspects that need to be considered, it is exposed the methods decision tree and Gradient Boosted Trees Model. This paper aims to propose a more practical approach for making more accurate stock predictions. The dataset including the stock bazaar values from the prior year has been considered first. The dataset was optimized for actual analysis through pre-processing. Therefore, the preprocessing of the raw dataset will also be the main emphasis of this work. Again, decision trees and gradient tree models are used on the pre-processed data set and the results thus obtained are analyzed. In addition, forecasting papers also address issues related to the usage of forecasting systems in actual situations and the correctness of certain normal value. This paper also presents a machine learning model for predicting stock stability in financial markets. Successful stock price forecasting greatly benefits stock market organizations and provides real solution to problems faced by investors.
Authors - Aneesh Kaleru, Chaitanya Neredumalli, Mrudul Reddy, Ramakrishna Kolikipogu Abstract - One major risk factor that contributes to traffic accidents globally is poor visibility in foggy situations. Drivers are seriously threatened by fog because it weakens contrast, hides important objects, and makes lane markings almost invisible. Recent developments in visibility enhancement methods for foggy circumstances are summarized in this paper, with a focus on picture defogging combined with object detection and lane aid. We analyze the application of models such as Conditional Generative Adversarial Networks (cGANs), Single Shot Multibox Detectors (SSD), All-in-One Defogging Network (AODNet), and You Only Look Once (YOLO) from the perspective of deep learning and computer vision. These methods have the potential to increase driver safety in inclement weather by identifying impediments, improving visibility, and offering lane guidance. The review also covers the limitations of these solutions, such as computational demands and requirements for real-time processing. Our goal is to provide researchers and practitioners with a comprehensive understanding of the current methods and their uses, allowing for the development of effective visibility enhancement systems that can prevent accidents and save lives.
Authors - Sindhu C, Taruni Mamidipaka, Yoga Sreedhar Reddy Kakanuru, Summia Parveen, Saradha S Abstract - India is a country with very rich ancient historical legacy. It preserved vast cultural and linguistic knowledge through stone inscriptions. Extracting text from ancient stone inscriptions and translating it into a language which is understandable by everyone is a very challenging task due to script variations, natural wear, and the uneven degraded surfaces of stone carvings. Our idea is to build a model which can extract the text from these stone inscriptions which were written in Telugu language and translate them into other Indian local languages. The Region-Based Convolutional Neural Network (R-CNN) model which is integrated with Tesseract OCR is trained on a custom dataset of 30,000 labelled images of Telugu script, encompassing Achulu (vowels), Hallulu (consonants), and Vathulu. By achieving a 96% accuracy in character detection, this model demonstrates significant reliability in recognizing Telugu characters from degraded and complex inscriptions. Data augmentation techniques, including rotations, flips, and shifts were used to further enhance the model’s robustness to different orientations and environmental conditions encountered in historical artifacts. The text which is extracted from the image is ultimately translated into Indian local languages using an API-based translation module, enabling a seamless interpretation of ancient content. This research contributes a comprehensive and automated solution for cultural preservation, giving us a scalable method to digitize and make historical inscriptions accessible to everyone which are integral to Telugu heritage and linguistic history.
Authors - Nisarg Dobariya, Rutik Dobariya, Rikita Chokshi, Sarita Thummar Abstract - The transition from traditional to smart grids has been driven by the pursuit of greater efficiency, reliability, and consumer engagement. While smart grids offer numerous benefits, they are vulnerable to cybersecurity threats. Intrusion detection systems (IDS) are indispensable tools for safeguarding smart grid operations by identifying and preventing malicious attacks. This research investigates the application of various IDS models, classifiers, datasets, and algorithms in smart grid environments. The study underscores the importance of using datasets specifically designed for smart grid networks to ensure accurate and reliable IDS performance. Moreover, the research demonstrates the potential of distributed approaches and advanced algorithms in enhancing IDS capabilities, thereby bolstering the security and resilience of smart grid infrastructure.
Authors - Revathy P, Rakshana A, Tinu A V, Vijayakumar R Abstract - The increasing demand for efficient package delivery has led to a challenge in detecting food spills during transit. Traditional methods rely on manual inspection, which is time-consuming and prone to human error. This study proposes an AI-based approach using Convolutional Neural Networks (CNNs) implemented with TensorFlow to detect both damaged packages and spilled food packets. The model is trained on a large dataset of package and food packet images, learning key features indicative of physical damage and identifying food spills. By fine-tuning pre-trained CNN architectures, the model achieves high accuracy in detecting both damage and spills. The interface is attached with an alert mechanism that notifies when damage or spill is detected. The TensorFlow framework is used for building, training, and deploying the model efficiently. The proposed system aims to automate package and food packet inspection, reduce human labor, and improve delivery service reliability by providing fast and accurate damage and spill detection.
Authors - Smita Mahajan, Archana Chaudhari, Ameysingh Bayas, Devika Shrouti Abstract - Drowsiness is a critical issue that contributes to a significant number of accidents in various scenarios, such as driving and hazardous work environments. Existing drowsiness detection projects often rely on subjective measures and single modality detection, leading to limited accuracy and applicability. This research proposes a drowsiness detection system that employs deep neural networks and machine learning-based object detection techniques to overcome these limitations. The ability of the recent drowsiness detection systems to reliably and impartially detect drowsiness is restricted. The proposed model uses computer vision and machine learning algorithms to identify drivers' drowsiness based on facial attributes like eye movement monitoring. The model aims to improve the accuracy and reliability of drowsiness detection by combining multiple modalities. The implementation includes using the Keras library, which is required for a Convolutional Neural Network (CNN) architecture. The model is trained on a customized dataset of facial images with open or closed eyes labels. The CNN discovers the complex relationships and features from the data, classifying drowsiness critically. The proposed drowsiness detection system's results demonstrate an optimistic accuracy of 98.88%. The system signals real-time alerts when the drowsiness in the behavior of the driver is caught, potentially averting accidents and enhancing safety. This technique suggests an accurate and trustworthy approach for detecting drowsiness in different domains, including driving and unsafe work environments, with 98.88% accuracy. This system can be a valuable means for improving safety and controlling the accidents caused by driver drowsiness.
Authors - Aswathkrishna S D, M. Sujithra Abstract - In today's rapidly evolving world, recognizing student food choices is crucial. This study explores food choices and how they align with areas containing restaurants and grocery stores. Clustering techniques including K-Means, Hierarchical Clustering, and DBSCAN were employed with the silhouette score used to validate and determine the most effective method for analysis. Based on food choices data sourced from Kaggle and location data from the Foursquare API, the research provides location recommendations for students. Suggestions guide students to areas that align with their food choices aiming to enhance their overall experience.
Authors - Ofaletse Mphale, V. Lakshmi Narasimhan, S. Sasikumaran Abstract - The Internet of Medical Things (IoMT) presents transformative potential for healthcare by enabling real-time patient monitoring, advanced diagnostics, and personalized treatments. However, its adoption in developing countries is hindered by significant ethical, security, and privacy challenges. Studies focused on developing countries often identify these challenges but rarely propose rigorous frameworks for successful adoption. This study employs a desktop search methodology to comprehensively review the existing literature, identifying crucial ethical, security, and privacy concerns associated with the IoMT adoption. Through this analysis, the study proposes potential mitigation strategies and a framework to facilitate the effective adoption of IoMT in developing countries. Findings will support healthcare decision-makers and policymakers in developing countries, enabling them to devise strategies that ensure ethical practices, secure patient data, and safeguard privacy in healthcare IoT integration. This will lead to improved healthcare delivery and enhanced patient outcomes.
Authors - Priyal Donda, Vatsal Upadhyay, Janhavi Gulabani, Sharvari Patil, Vinaya Sawant Abstract - Phishing is increasingly being one of the frequent cyber-attacks. Since this trend has seen the incidence increased significantly in the last few years, people and organizations have been highly affected by data breaches and financial losses. Such growth only increases the demand for effective mechanisms of defense, as traditional approaches of machine learning like SVM, Random Forest, and Long Short-Term Memory networks often fail to detect phishing attempts with accuracy. SVMs can be computationally expensive, sensitive to noise, and require careful selection of kernel functions, while LSTMs are complex, prone to overfitting, and require substantial amounts of labeled data. In light of these limitations, the use of GANs has been recent in order to improve detection capabilities. GANs create realistic phishing URLs that advanced detection models struggle to distinguish, using semi-supervised training to differentiate between adversarial and legitimate URLs. Specifically, this holistic approach grapples with the sophistication of phishing attacks and places an emphasis on adaptive defense, since it has changed the basis for detection from content-based to URL-based techniques. Finally, these novel approaches introduce a promising pathway for the mitigation of phishing risks and sensitive information safeguarding, thus building security strength in the digital world.
Authors - Amol Mashankar, Smita Kalokar Abstract - The retail business marketplace is experiencing a significant shift, with a growing emphasis on the innovations brought by internet of things (IoT) technology. The retail aspect is rapidly evolving, driven by new improvements in internet technology, which play a important role in the transformation of the retail sector. The new updation involves continuously adapting to the fast-paced changes within the retail environment. New techniques and innovations are emerging daily to better address customer needs and satisfaction preferences. This paper focus on to explore the practices and performance of IoT technology in the retail sector. It also emphasis an analytical framework for evaluating the approaches to IoT technology practices and their effectiveness in retail stores.
Authors - Bakka Vamshi, Munnuru Umakanth, Kadwasra Swapna, Punuru Venkata Usha Sree, Mannepalli Rohini Sri, Sushama Rani Dutta Abstract - Predicting heavy rainfall remains a significant challenge for meteorological departments as it greatly impacts economies and human lives. Severe rainfall can result in natural disasters like floods and droughts, impacting millions of people globally every year. Precise rainfall prediction is especially important for nations like India, where agriculture serves as a key economic pillar. Due to the atmosphere’s dynamic nature, statistical methods often fall short in achieving high prediction accuracy. The complex, nonlinear characteristics of rainfall data make Artificial Neural Networks a more effective method. This paper reviews and compares various methods and algorithms employed by researchers for rainfall forecasting, presenting the findings in a tabular format to make these techniques accessible to non-specialists.
Authors - Rajitha Kotoju, Sugamya Katta, Abrar Khan Abstract - Real-time air quality monitoring and predictive pollution control are critical for addressing escalating environmental and public health challenges, particularly in low-income areas with limited infrastructure. This paper explores the integration of Big Data analytics and IoT to develop cost-effective, scalable solutions for real-time air quality assessment. The proposed framework aims to identify pollution patterns, predict air quality trends, and provide actionable insights for policymakers. A unique feature of this study is its emphasis on low-cost sensor deployment and edge-computing techniques to ensure accessibility in resource-constrained settings. The interdisciplinary approach combines environmental science, AI, and public health perspectives to establish a holistic framework for data collection, analysis, and decision-making. Additionally, this paper addresses the integration of findings into policy frameworks by proposing data-driven recommendations for urban planning, industrial regulation, and community health interventions. The results demonstrate significant advancements in predictive accuracy and actionable intelligence generation while minimizing implementation costs.
Authors - D.K. Chaturvedi, Nisha Verma Abstract - Artificial Intelligence and Machine Learning (AIML) are quickly proceeding in many areas. These technologies, including the smart footwear (SF) industry, have significantly impacted the consumer goods market. AIML are widely used in the design and production of SF. There are different applications of SF such as healthcare SF, assistive SF for old age persons or impairments, navigating footwear for unknown areas, mobility and gait analysis, safety footwear, anti-skid footwear, footwear for army personnel, and power generated footwear etc. The SF helps in acquisition of real-time data of patients to monitor and suggest suitable treatment. Besides these, SF can be classified based on the different architecture and processing techniques. This paper includes different research studies conducted in the past on various tools and techniques used to create smart footwear for different applications.
Authors - Kajal Joseph, Deepa Parasar Abstract - This study conducts a predictive analysis of company status using various machine learning algorithms, aiming to identify the models that deliver the highest accuracy and reliability for decision-making in finance and business intelligence. The study employs a range of algorithms, including Logistic Regression, DecisionTreeClassifier, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naive Bayes, Gradient Boosting Machines (GBM), XGBoost, AdaBoost, LightGBM, CatBoost, and Extra Trees Model, each rigorously tested on a preprocessed dataset split into training and testing sets to ensure robust validation. (Kunjir et al., 2020) Results indicate that ensemble models, particularly XGBoost and Random Forest, outperformed other methods, achieving accuracy rates exceeding 93%. This high level of performance highlights the value of ensemble techniques for handling complex predictive tasks, showcasing their suitability for applications where precise forecasting is critical. The study underscores the importance of model selection in predictive analytics, as it directly impacts the reliability of predictions in financial contexts. These findings suggest that machine learning, especially ensemble models like XGBoost and Random Forest, can significantly improve the accuracy of company status predictions, offering a dependable tool for stakeholders operating in uncertain environments. This research contributes valuable insights into the efficacy of machine learning in predictive tasks, advocating for data-driven decision-making approaches that can enhance business intelligence and strategic planning. (Liaw et al., 2019)
Authors - Meenu Suresh, Tonny Binoy, Saritha M S, Vimal Babu P, Dheeraj N, Aiswarya R Lakshmi Abstract - The present work introduces a video steganography technique which employs Finite Ridgelet Transform (FRT) and Elliptic Curve Cryptography (ECC)-ChaCha20 encryption to hide confidential information. The proposed method begins by identifying key frames through the detection of scene changes. The FRT is employed to analyze the key frames, extracting their orientation and subbandswithin which the secret data is encoded. To boost security, ECC-ChaCha20 encryption technique serves as a preprocessing step prior to incorporating the secret data. The technique attains an embedding capacity of 72%, SSIM of 0.9890 and PSNR value range from 70dB and 72 dB. The experimental results highlight that the algorithm besidesboosting security also ensures superior resilienceand video quality.
Authors - Jagriti Singh Chundawat, Ashish Kumar, Monika Saini Abstract - The purpose of this paper is to optimize the availability of a thermal power plant. A thermal power plant (TPP) is a comprehensive system with multiple interconnected subsystems which are used for power generation. This TPP system has three subsystems such boiler, superheater and reheater. These subsystems connect to each other in series configuration. To improve the availability of the system a study-state availability is derived with the help of normalizing equations and the chapman Kolmogorov equations are derived from Markov birth-death process. The system’s failure and repair rates are statistically independent and exponentially distributed. The numerical results show that availability increases from 0.997903 to 0.998725 as the repair rate increases.
Authors - Mannem Sri Nishma, Satendra Gupta, Tapas Saini, Harshada Suryawanshi, Anoop Kumar Abstract - Face recognition-based authentication has become a critical component in today's digital landscape, particularly as most business activities transition to online platforms. This is especially evident in the finance and banking sectors, which have shown significant interest in adopting online processes. By leveraging this technology, these industries can enhance operational efficiency, promote business growth, reduce reliance on manpower, and automate several processes effectively. However, face recognition systems are susceptible to face spoofing attacks, where malicious actors can attempt to deceive these systems using facial images or videos. Some attackers even use masks resembling authorized individuals to trick recognition cameras into perceiving them as real users. To counter such threats, liveness detection has emerged as a critical research area, focusing on identifying and preventing face spoofing attempts. The proposed approach utilizes a deep learning technique tailored for face liveness detection. The experiments are conducted using the Replay-Mobile, MSU-MFSD, Casia-FASD and our own datasets, which are widely used for recognizing live and spoofed faces. The system achieved an impressive area under the ROC curve (AUC) of 0.99, demonstrating its effectiveness in detecting face spoofing.
Authors - Kusuma B S, Meghana Murthy B V, Preksha R, Srushti M P, C Balarengadurai Abstract - Against the backdrop of either a Deaf World or hearing people, the major challenges which face modern society concern communication barriers in general. The paper proposes a system for translation through gestures in Indian Sign Language to audio and video outputs for non-signers to enable easy interaction with them. Advanced machine learning techniques, such as Support Vector Machine and Convolutional Neural Network, will be used to enable this tool to recognize motions of ISL in real time. It converts these into the correct format for video and audio. In this respect, the paper claims to "make communication more accessible and bridge the gap in communication in which gestures are recognized and translated." Real-time recognition algorithms overcome the challenges faced by hand gesture detection to provide an intuitive and seamless interaction experience. This approach is an effective strategy to enhance communications in government and industry with special focus on smart writing. Results confirm this method's promise in the broader social interaction by significantly improving the speed and accuracy of deaf individuals.
Authors - Bitan Pratihar Abstract - We, human-beings, have two different forms of memory, namely pulling memory and pushing memory (also known as working memory). A pure pulling memory pulls a person towards itself, and consequently, he/she spends some significant amount of time on memorizing the incident but does not gain anything significant in his/her decision making directly. On the other hand, a pure pushing memory pushes a human-being to take some decisions, and thus, it may have direct influence on his/her learning. However, neither pure pulling memory nor pure pushing memory alone may be beneficial to effective learning of human brain. A proper combination of pulling and pushing memories may be required to ensure a significant effect of memory on learning of neural networks. The novelty of this study lies with the fact of formulating it as an optimization problem and solving the same using a recently proposed nature-inspired intelligent optimization tool. The effectiveness of this novel idea of correlating the combined form of memory with learning of neural networks has been demonstrated on two well-known data sets. This combined form of memories is found to have a significant influence on learning of neural networks, and this proposed approach may have the potential to solve the well-known memory loss problem of neural networks.
Authors - Martin Mollay, Deepak Sharma, Pankajkumar Anawade, Chetan Parlikar Abstract - The primary research intention of the present study is to find out the impacts of laws such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) on the digital marketing landscape. The set of regulations relates to data protection, which involves a stringent regime for how the firms gather, process, and hold the privacy of their clients. Therefore, the two main bottlenecks of marketers are fewer consent mechanisms, less data, and a need for more options for personalizing. Still, technology is a fashionable thing that has been launched, and the effectiveness of the new technology revitalizes it. Firms mostly turn to first-party data that question the need for intermediaries. This means that they can collect information directly from the consumer, which then naturally results in much more productive and meaningful customer relation-ships. Getting hold of advanced technologies, for instance, artificial intelligence and machine learning, which work with smaller datasets, also provides a window for companies to discover a large number of customized, and possibly even more valuable, aspects through customer behavior without invading the privacy of the person who is identical to the threat of the law. Additionally, the price of compliance with the regulations is high, notably for Small and Medium sized Enterprises (SMEs). In contrast, it is the most highly cost-effective way for the consumer to win consumers’ trust in the brand and make them loyal to it in the long run. In this new era, ethical marketing follows the footsteps of the evolutionary journey where complete openness and consumers’ private space value are the main topics. Personal data can be acquired in a way that is not compliant with privacy laws. However, zero-party data or consumer information given to businesses might still be the source for personalized experiences that are privacy compliant.
Authors - Sachi Joshi, Upesh Patel Abstract - Cancer is a grave category of illnesses in which the body's aberrant cells proliferate and spread uncontrollably. It can appear in nearly every tissue or organ and take many different forms, each with its own distinct set of symptoms and side effects. Environmental variables, lifestyle decisions, and genetic abnormalities are typically linked to the development of cancer. The varied approaches to cancer diagnosis are examined in this study, with a focus on early detection and therapeutic strategies. This literature review covers a wide range of cancer kinds, such as brain tumours, leukaemia, breast, lung, and cervical cancer, and offers recommendations for creating reliable ma-chine learning-enhanced cancer detection techniques. The research elucidates several applications, techniques, and comparative analysis in this significant subject, ranging from imaging analysis to biomarker identification. The study explores the developing methods that lead to a more precise diagnosis. The study offers insights with a thorough examination of the benefits, drawbacks, and innovations of each technique, ranging from conventional diagnostic procedures to state-of-the-art technologies. It also directs future research efforts towards the hunt for more effective personalized illness management.
Authors - Akshay Honnavalli, Hrishi Preetham G L, Aditya Rao, Preethi P Abstract - In todays information-driven world, organizing vast amounts of textual data is crucial. Topic modelling, a subfield of NLP, enables the discovery of thematic structures in large text corpora, summarizing and categorizing documents by identifying prevalent topics. For Hindi speakers, adapting topic modelling methods used for English texts to Hindi is beneficial, as much of the research has focused primarily on English. This research addresses this gap by focusing on Hindi language topic modelling using a news category dataset, providing a comparative analysis between traditional approaches like LDA, LSA, NMF and BERT-based approaches. In this study, six open-source embedding models supporting Hindi were evaluated. Among these, the l3cube-pune/hindi-sentence-similarity-sbert model exhibited strong performance, achieving coherence scores of 0.783 and 0.797 for N-gram (1,1) and N-gram (1,2), respectively. Average coherence scores of all embedding models significantly exceeded traditional models, highlighting the potential of embedding models for Hindi topic modelling. Also, this research introduces a novel method to assign meaningful category labels to discovered topics by using dataset annotations, enhancing the interpretation of topic clusters. The findings illustrate both the strengths and areas for improvement in adapting these models to better capture the nuances of Hindi texts.
Authors - Guhan Senthil Sambandam, Priyadarshini J Abstract - Machine learning has significantly impacted daily life, with machine translation emerging as a rapidly advancing domain. In healthcare, machine learning presents opportunities for innovation, particularly in translating medical documents into low-resource languages like Tamil. This research develops a transformer-based model fine-tuned for medical terminology translation from English to Tamil. A major challenge was the lack of English-Tamil medical datasets, addressed through innovative data collection methods, such as extracting bilingual subtitles from Tamil YouTube videos. These datasets complement existing resources to enhance model performance. The final model was deployed as a REST API using a Flask-based server, integrated into a React Native mobile application. The app enables users to scan English medical documents, extract text via on-device Optical Character Recognition (OCR), and obtain Tamil translations. By combining advanced Natural Language Processing (NLP) techniques with user-friendly application design, this end-to-end system bridges linguistic gaps in healthcare, providing Tamil-speaking populations with improved access to critical medical information. This study highlights the potential of NLP-driven solutions to address healthcare disparities and demonstrates the feasibility of adapting machine translation systems to specialized domains with resource limitations. The approach also emphasizes scalability for broader applications in similar low-resource settings.
Authors - Manohar R, N Abhishek, Nagesh S, Sumith R, C Balarengadurai Abstract - Water quality monitoring is essential for public health and environmental stewardship. Conventional methods, while effective, are often costly, time-intensive, and require specialized skills. In response to these limitations, this paper explores machine learning as a rapid, scalable solution to classify water quality using key parameters, including pH, turbidity, organic carbon, and contaminants. By implementing algorithms such as Random Forest, SVM, and other advanced models, we seek to enhance the precision of water purity assessments. This paper shows the potential of ML applications in real-time monitoring, addressing the need for accessible, cost-efficient, and accurate water quality solutions suitable for broad deployment across diverse environments.
Authors - Hrudai Aditya Dharmala, Ajay Kumar Thallada, Kovvur Ram Mohan Rao Abstract - Recent advances in vision-language models have demonstrated remarkable multimodal generation capabilities. However, their typical reliance on training large models on massive datasets poses challenges in terms of data and computational resources. Drawing inspiration from the expert-based architecture of Prismer, we propose a novel framework for contextual visual question answering specifically tailored to remote sensing imagery. Our methodology extends the Prismer architecture through a two-stage approach: first, by incorporating a domain-specific segmentation expert trained on remote sensing datasets, and second, by integrating a fine-tuned Large Language Model (Mistral 7B) optimized using Parameter-Efficient Fine-Tuning (PEFT) with QLoRA for remote sensing terminology, with hyperparameters optimized with help of Unsloth framework. The segmentation expert performs the analysis of remote sensing imagery, At the same time, the language model acts as a reasoning expert, combining domain-specific knowledge with natural language understanding to process visual contexts and generate accurate responses. In our framework, the use of the Unsloth fine-tuning approach for the language model helps maintain high performance within the defined scope of remote sensing classes and terminology while avoiding hallucination or deviation from established classification schemas. This opens an exciting direction for making the use of Earth observation data more accessible to end-users, demonstrating significant improvements in accuracy and reliability compared to traditional approaches. Experimental results validate that this architecture effectively balances domain expertise with computational efficiency, providing a practical solution for remote sensing visual question answering that requires substantially fewer computational resources compared to end-to-end training of massive models.
Authors - Jenat Arshad, Afruja Akter, Tanjina Akter, Kingkar Prosad Ghosh, Anupam Singha Abstract - Cryptocurrencies have emerged as a significant financial asset class, attracting global attention for their potential to disrupt traditional financial systems. Due to its extreme price volatility and ability to be traded without the assistance of a third party, cryptocurrencies have gained popularity among a wide range of individuals. This paper presents a comprehensive study of machine learning techniques, particularly deep learning models such as Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and Artificial Neural Network (ANN) in predicting cryptocurrency prices. Furthermore, this study addresses the security and privacy challenges inherent to blockchain technology, upon which cryptocurrencies operate. We predict the prices of popular cryptocurrencies like Bitcoin, and Ethereum, and lesser-known ones like Binancecoin, Litecoin, and Ripple through a hybrid deep learning model. This paper also compares cryptocurrency price prediction with machine learning models like GRU, ANN, and our proposed model Hybrid LSTM-GRU. The results demonstrate the efficacy of machine learning in price prediction, highlighting blockchain's potential to enhance security and privacy in financial transactions. Our model gives the value of MSE, RMSE, MAE and MAPE to determine the forecasting. We’ve also added the manual calculation for each metric and compared the actual price with the predicted price that our model gave.
Authors - Harsha S Khurana, Payal D Joshi Abstract - Stress is a major health concern that significantly affects mental stability and can have adverse effects on physical well-being if prolonged. Early detection of stress can help and prevent health-related issues. Individual stress patterns are detected using a variety of bio-signals, including thermal, electrical, auditory, and visual cues which are invasive methods. But according to the well-known saying statement, "Face is a mirror of mind," one can observe one’s emotion or mental state on one’s face. Based on this, Investigated the potential of using facial expressions as a non-invasive method to detect stress levels. Facial expressions could be analyzed and classified as stress and non-stress by examining facial expressions. To solve this problem we have used pretrained network models - Inception, Xception, MobileNetv2, Vgg19, EfficientNet deep learning models, and Affectnet Dataset for stress detection and also represent the comparative study of networks based on confusion and performance metrics. Testing on a separate set of data of images indicates that the MobileNetv2 and Xception models give more accuracy for stress detection.
Authors - Rinkesh N Parmar, Payal D Joshi Abstract - Cotton, an essential crop for the textile industry and millions of farmers, is vulnerable to diseases that can significantly affect yields and profitability. Traditional methods of disease detection, relying on expert visual inspections, are labour-intensive, time-consuming, and prone to errors, often causing delays in addressing problems. This study investigates the use of Convolutional Neural Networks (CNNs) for automated, early, and accurate detection of cotton diseases. CNNs are effective at extracting hierarchical features from raw image data, making them ideal for image classification tasks. In this approach, a labelled dataset of cotton plant images is utilized to train the CNN model, incorporating data augmentation to enhance variability and generalization. The model employs convolutional layers for feature extraction, max-pooling layers for dimensionality reduction, dropout layers for regularization, and fully connected layers for classification. The Adam optimizer, known for faster convergence, is used during training, along with categorical cross-entropy loss. The evaluation is based on accuracy, precision, recall, and F1-score. The model showed significant improvements in performance. The baseline CNN achieved 92.34% accuracy, but advanced architectures like Hybrid CNN-LSTM, DenseNet-121, ResNet-50, and InceptionV3 enhanced accuracy by 2-3%, along with increased precision, recall, and F1-score. The Hybrid CNN-LSTM model outperformed others, achieving 94.5% accuracy, 93.5% precision, 93.2% recall, and 93.3% F1-score. These results suggest that CNN-based models, particularly Hybrid CNN-LSTM, offer substantial improvements in cotton disease detection. The incorporation of data augmentation and dropout regularization strengthens the model, making it effective for real-time agricultural disease management. Future work will focus on expanding the dataset, improving the model, and implementing it in real-world cotton farming practices.
Authors - Joven A. Tolentino Abstract - The growing demand for electricity necessitates effective monitoring and forecasting of consumption trends. This study employs ARIMA modeling, using data from the Department of Energy, Philippines, to analyze and predict electricity consumption. The forecast for the next two years indicated an 18.99% increase in consumption between 2016 and 2017.To enhance analysis, the predicted data was clustered using the K-Means algorithm to group months with similar consumption patterns. This approach identified periods of high, medium, and low electricity usage, providing valuable insights into peak demand months. Such data-driven findings can guide electricity providers in prioritizing resources and implementing strategies to address fluctuations in consumer demand effectively. This study emphasizes the importance of forecasting and clustering as tools for decision-making to mitigate challenges arising from increasing electricity demand.
Authors - Manmeet Borkar, Suneeta Raykar Abstract - Monitoring biomarkers is essential for patients to effectively manage their health profiles and prevent potential complications. Regular tracking of these indicators allows for timely interventions and better control over one’s health, particularly when the methods employed are non-invasive and grant convenience and comfort to the patient. Conventionally, this monitoring is accomplished in pathology laboratories, by collecting blood samples or finger-pricking, which can be distressing and impractical for regular use. Given the growing need for more accessible and affordable healthcare solutions, the development of a cost-effective non-invasive method has become crucial. We propose the use of machine learning models to enable non-invasive measurement of biomarkers such as Total Cholesterol, Uric acid and Blood Sugar. Several Machine learning algorithms, including Linear Regression, K-Nearest Neighbors (KNN), Decision Tree, Random Forest and Support Vector Regression (SVR), were applied to the datasets constructed using the MAX30102 sensor. The metrics used to evaluate regression models were Mean Square Error (MSE) and Coefficient of determination (R²). The final prediction model was built using the algorithm that yielded the highest Coefficient of determination (R²). A user-friendly interface was developed using Tkinter, allowing the input of sensor values from the MAX30102 sensor. The prediction of biomarker values promotes health awareness and timely alerts against potential complications. The results obtained using this approach were validated against laboratory blood reports, revealing an average offset of less than 10% in the predictions.
Authors - Kirthika. P, M. Suresh, S. Kanagaraj Abstract - This paper explores the social barriers faced by Indian plantation communities. It focuses on how these obstacles impact their well-being, productivity, and social mobility. By analyzing historical, socioeconomic, and cultural factors, the study uncovers the multifaceted challenges plantation workers encounter, including income, education, social position in the community, social networks, migration, exploitation, and working and living conditions. The DEMATEL approach identifies the barriers and analyzes the interrelationships among those that impact social barriers among plantation workers. This paper identified seven barriers of impact from a literature review followed by interviews with experts to interpret the interconnection of barriers and investigate the interrelationships. The result says that income and education are the key barriers impacting the lives of plantation workers in their society. The present study incorporates the DEMATEL approach model to analyze the critical barriers in mapping the social barriers of plantation workers. The DEMATEL approach model is the first attempt to study the interrelationship among the barriers. The research overviews the prevailing issues through field surveys, interviews, and literature reviews. The paper will conclude with actionable recommendations aimed at policymakers, community leaders, and stakeholders to mitigate these barriers and promote a more inclusive and equitable environment for plantation workers.
Authors - Aishani Das, Sobitha Ahila, Sreyashi Dey Abstract - Sentiment analysis within the food industry offers essential insights into customer satisfaction, product perception, and emerging concerns. A novel sentiment classification model is developed for Amazon food reviews, leveraging Sentiments are categorized as positive, neutral, or negative using techniques from Natural Language Processing and Machine Learning. Traditional ML algorithms, such as Logistic Regression, Naive Bayes, and Support Vector Machines, are combined with the BERT deep learning model to enhance classification accuracy. With a dataset of over 500,000 reviews sourced from Kaggle, the methodology includes data cleaning, feature extraction, exploratory data analysis, model training, and evaluation. Initial findings demonstrate SVM’s high predictive accuracy in sentiment classification, while BERT’s advanced contextual understanding suggests further enhancements. Applications of this model extend to real-time feedback systems that assist businesses in identifying and addressing customer sentiments promptly. Future developments aim to improve accuracy, incorporate a diverse range of datasets, and integrate real-time processing and multilingual analysis for broader, more effective sentiment analysis capabilities.
Authors - Tinoy Santra, Sahil Neekhra, Ritik Gupta, Gunabalan Ramachandiran Abstract - With the rising environmental degradation and increasing global warming, electric vehicles are the promising concept in the automobile industry. Different sources of energy are available for giving power to drive the vehicle. Sunlight being an efficient and abundant resource, the world is moving towards solar energy leaving behind conventional power resources. Moreover, battery based electric vehicles have short driving range and speed which is not acceptable in the dog-eat-dog market. This paper discusses a simple approach for BLDC motor driven electric vehicle powered by buck-boost converter. The primary energy source is solar energy, and the PI controller holds the DC-DC converter's output constant. A 660 W, 48 V BLDC motor driven electric bike system is worth an alternative when it is solar powered which solves utmost all the problems faced in usage of EVs. The circuit is simulated in MATLAB environment and output parameters are observed for different load conditions. Overall, the motive is to prove that electric vehicles are more efficient and cost effective than the conventional ones.
Authors - Shruti Anghan, Tirth Chaklasiya, Priyanka Patel Abstract - Technology is an indispensable tool that many industries use to transcend and arrive at the best possible results. A very significant part of the Indian economy constitutes the agricultural sector. Half of the country's workforce is still employed by the agriculture industry. What plays a critical role in affecting the agricultural sector is the natural environment within which it operates, and it throws up many challenges in real farming operations. Most agricultural processes in the country have been old-fashioned and the industry is not ready to step into new technologies. Effective technology can enhance production and reduce the greatest barriers in the field. Today, farmers mostly plant crops not based on soil quality but the market value of the crop and what the crops can return to them. This might impact the nature of the land and the farmer also. Properly applied, modern technologies such as machine learning and deep learning can help revolutionize these industries. It shall show how to apply these technologies properly to give the farmer maximum support in the crop advice field.
Authors - Bimal Patel, Ravi Patel, Jalpesh Vasa, Mikin Patel Abstract - The study delves into Tableau's unique characteristics, including its intuitive interface, robust analytics capabilities, and advanced visualization features. By leveraging these features, Tableau empowers users to transform complex datasets into actionable insights, facilitating data-driven decision-making across various domains. The paper explores the extensive applications of Tableau in key industries such as finance, healthcare, retail, and education. In finance, Tableau aids in risk management and performance analysis, while in healthcare, it enhances patient care and operational efficiency through detailed data visualizations. The retail sector benefits from Tableau's ability to analyze sales performance and customer behavior, and in education, it tracks student performance and engagement metrics. Additionally, this research identifies and addresses common challenges associated with data visualization using Tableau, such as handling large datasets, ensuring data accuracy, and maintaining user engagement. The paper provides practical solutions and best practices to overcome these hurdles, ensuring optimal use of Tableau's capabilities. The paper shows how Tableau can be used to help different industries with their specific needs and problems using real-life examples. This study serves as a valuable resource for professionals and researchers seeking to maximize the potential of Tableau in their respective fields.
Authors - Aditi Zeminder, Vaibhav Patil, Prathamesh Raibhole, S V Gaikwad Abstract - This paper presents a part of the study of a collaborative robot (cobot) designed for optimization of work tasks, focusing on selection and workplace. This project investigates best practices by developing a kinematic editing library and using ROS and RViz to perform simulations to analyze and improve motion planning. Conducted an exhaustive review of the existing research literature on collaborative robot control and efficiency and will examine the usage of commercial collaborative software, such as Elephant Robotics' myCobot and Dobot, in introducing the interface design. The Kivy-based control interface was designed to allow users to effectively interact with the robots and adjust parameters to complete tasks. This paper provides an overview of the process adopted, the challenges encountered during development and initial testing, and lays the groundwork for future developments including hardware integration and additional kinematic optimization.
Authors - Eshwari Khurd, Shravani Kamthankar, Avani Kelkar, Ravinder B. Yerram Abstract - One of the major challenges encountered when it comes to speech recognition, medical imaging, and multimedia processing for radar or weather forecasting applications, is noise interference in audio and image signals that invariably affect algorithmic precision and dependability. Denoising is responsible for removing unwanted noise while keeping intact the necessary details in the signal. An effective denoising method for audio and image signals is under continuous research across multiple parameters taken into consideration giving priority to signal-to-noise ratio (SNR). In this paper, we have surveyed various such denoising methods with a focus on the ones using Principal Component Analysis (PCA) and Ensemble Empirical Mode Decomposition (EEMD).
Authors - Renuka Deshmukh, Babasaheb Jadhav, Srinivas Subbarao Pasumarti, Mittal Mohite Abstract - In response to the issue of growing garbage, researchers, foundations, and businesses worldwide developed concepts and created new technology that sped off the procedure. Trash comes from a variety of sources, including municipal solid trash (such as discarded food, paper, cardboard, plastics, and textiles) and industrial garbage (such as ashes, hazardous wastes, and materials used in building and demolition). Contemporary waste management methods often take sociological factors into account in addition to technological ones. This review paper's goal is to talk about the potential applications of cutting-edge digital technology in the waste disposal sector. With reference to smart cities, this study aims to comprehend the environment, including the opportunities, barriers, best practices at present, and catalysts and facilitators of Industry 4.0 technologies. An innovative approach for examining the use of digital technology in smart city transformation is put out in this study. Analysis of the suggested conceptual framework is done in light of research done in both developed and developing nations. The study offers case studies and digital technology applications in trash management. This article will examine the ways in which waste management firms are utilizing cutting-edge technology to transform waste management and contribute to the development of a healthier tomorrow.
Authors - Arvin Nooli, Preethi P Abstract - Recognizing emotional states from electroencephalogram, or Electroencephalogram (EEG), signal data is challenging due to its large dimension and intricate spatial dependencies. Our project illustrates a novel approach to Electroencephalogram (EEG) data analysis in emotion recognition tasks that employ Dynamic Graph Convolutional Neural Networks (DGCNN). Our novel architecture takes advantage of the inherent graph structure of Electroencephalogram (EEG) electrodes to effectively capture spatial relationships and dependencies. Our approach used a refined DGCNN model to process and classify the data into four primary emotional states- Happy, Sad, Fear, and Neutral, we configured the DGCNN with 20 input features per electrode, optimized across 62 electrodes, and utilized multi-layered graph convolutions. The model achieved an overall classification accuracy of 97%, with similarly high macro and weighted average scores for precision, recall, and F1-score, demonstrating its resilience and accuracy.
Authors - Chitraksh Madan Singh, Yash Kumar, Lakshya Gattani, A.Anilet Bala, Harisudha Kuresan Abstract - This study presents an analysis of Instagram reach using Passive Aggressive, Decision Tree, Random Forest, and Linear Regression models. The goal is to predict the impressions generated by posts based on features like likes, saves, comments, shares, profile visits, and follows. Using Instagram data, machine learning algorithms are applied to forecast the post reach, helping marketers optimize content strategies. Quantitative metrics such as Mean Squared Error (MSE) and R-squared (R2) are used to evaluate model performance, with Random Forest showing superior accuracy compared to other models.
Authors - Shreyas Shewalkar, Shweta Autade, Aditi Sonje, M.R. Kale Abstract - With the growing need for automated text recognition and image processing, we have explored techniques that enhance the accuracy of handwritten character recognition while simultaneously addressing image restoration challenges. Handwritten English Character Recognition leverages deep learning (DL) techniques to classify and accurately identify characters from scanned or photographed documents. A deep learning-based approach is employed to recognize the patterns in handwritten text, ensuring high precision in distinguishing between characters despite variances in writing styles. In addition to recognition, colorization of grayscale images has gained attention, where DL models predict and apply realistic colors to black and white images. The recognition process applies CNN (Convolutional Neural Networks) for character identification.
Authors - Sangjukta Halder, Renuka Deshmukh Abstract - This study scrutinizes the impact of ICT-driven financial literacy agendas in India, focusing on their role in promoting financial inclusion and enhancing governance. By leveraging digital tools such as mobile apps, online courses, and e-governance platforms, these programs have effectively increased financial literacy, particularly among underserved populations. The research highlights that while challenges such as the digital divide, language barriers, and varying levels of digital literacy persist, these programs significantly empower citizens to make conversant financial choices and participate more actively with public fiscal management. The incorporation of financial literateness into digital platforms also fosters greater transparency and accountability in governance. For the purpose of improving these programs, legislators, educators, and tech developers may benefit greatly from the insights this research offers. Additionally, it makes recommendations for future research topics to investigate the long-term effects of financial literacy programs powered by ICT on financial behaviours and governance in various socioeconomic situations across India.
Authors - Yasharth Sonar, Piyush Wajage, Khushi Sunke, Anagha Bidkar Abstract - Emotion recognition from speech is a crucial part of human-computer interaction and has applications in entertainment, healthcare, and customer service. This work presents a speech emotion recognition system that integrates machine learning and deep learning techniques. The system processes speech data using Mel Frequency Cepstral Coefficients (MFCC), Chroma, and Mel Spectrogram properties that were extracted from the RAVDESS dataset. A variety of classifiers are employed, including neural network-based multi-layer percept, Random Forest, Decision Trees, Support Vector Machine, and other traditional machine learning models. We have created a hybrid deep learning system to record speech signals' temporal and spatial components. a hybrid model that combines convolutional neural networks (CNN) with long short-term memory (LSTM) networks. With an accuracy of identifying eight emotions—neutral, calm, furious, afraid, happy, sad, disgusted, and surprised—the CNN-LSTM model outperformed the others. This study demonstrates how well deep learning and conventional approaches may be used to recognize speech emotions.
Authors - Satya Kiranmai Tadepalli, Sujith Kumar Akkanapelli, Sree Harsha Deshamoni, Pranav Bingi Abstract - This paper in detail analyzes how generative AI and encoder-based architectures are drastically changing the realm of video generation with multimodal inputs such as images and text. The application of CNNs, RNNs, and Transformers so neatly serves to encode divergent modalities that blend into the seamless synthesis of realistic video sequences. It is based on the up-and-coming fields of generative models like GANs and VAEs, in bridging from static images to video generation. However, this represents a significant leap forward in the technology of video creation. It also goes into great detail on the complexities of multimodal input, working to balance coherence over time as well as semantic alignment of what's being produced. In the above-described context, it can now be realized how the role of encoders translates visual and textual information into actionable representations for generating video. What follows is a survey on recent progress in adopting Generative AI and multimodal encoders, discussions on what challenges are encountered today, and possible future directions that ultimately lay emphasis on their potential to assist video-related tasks and change the multimedia and AI communities.
Authors - Vathsal Tammewar, Bharat Sharma, Dharti Sorte, R. Sreemathy Abstract - Accelerated facial aging using GANs has been the key interest area in generative modeling and facial analysis fields, which offers significant breakthrough in age progression and regression solutions. This survey conducted an extensive review on techniques of GAN- based approaches for accelerated facial aging, emphasizing highly realistic and controllable aging transformations. Many of these methods applied in forensic investigations, entertainment industries, or age-invariant facial recognition systems, which are vivid demonstrations of the versatility and practical relevance of such methods. While such recent breakthroughs hold great promises, several issues remain; namely high-fidelity transformations to preserve important facial details do not fully diminish biases due to imbalanced datasets, and temporal consistency when age progressions or regressions consist of sequential ages is also critical. Computational efficiency and real-time applicability are still the most critical areas of focus. This paper probes into the strengths, limitations, and open challenges of existing approaches, while emphasizing the importance of innovations such as improved loss functions, diverse and representative training datasets, and hybrid architectures. Thus, this survey contributes to synthesizing current progress and outlining future research directions for advancing the field of GAN-based facial aging technologies.
Authors - Khushi Mantri, Abhishek Masne, Shruti Patil, Girish Mundada Abstract - In medical diagnostics, identifying bone fractures is a crucial task that is traditionally dependent on radiologists deciphering X-ray pictures. However, human factors like experience or exhaustion can occasionally cause delays or inaccuracies in diagnosis. The construction of an automated system for bone fracture identification utilizing Convolutional Neural Networks (CNN), a deep learning method that performs especially well in picture processing, is examined in this research. With the use of a tagged dataset of X-ray pictures, the suggested method can efficiently and accurately detect fractures. Prior to feature extraction using CNN layers which are trained to distinguish between fractured and non-fractured bones the images are pre-processed to improve clarity. In order to assist medical practitioners in making prompt, correct judgments, the final classification attempts to increase diagnostic accuracy while decreasing the amount of time needed for analysis. The potential of incorporating machine learning into healthcare to lower diagnostic errors and enhance patient outcomes is also discussed in this overview paper, which includes examines recent developments in CNN-based medical picture categorization.
Authors - Shradha Jain, Sneha Suman, Insha Khan, Ashwani Kumar, Surbhi Sharma Abstract - With continual advancements in deep learning, the potential misuse of deep fake is increasing and its detection is in a major scope of work. A model is trained to recognize the patterns in input data, deep fake recognize those patterns in a fabricated way. Sometimes a small, intentional change is added in the data points, these changes are undetectable to humans and confuse the learning model. Those changes are called adversarial perturbations. Compressive adversarial perturbations aim to make those changes even smaller and harder to detect. Authors explore a sophisticated framework - ComPAD (Compressive Adversarial Perturbations and Detection) which is used to detect adversarial attacks. This paper explores the strategies, and provides comparative analysis of methods used by different researchers. Various datasets including UADFV, DeepfakeTIMIT, LFW, FF++, and Deeperforensics are evaluated to achieve the highest metrics. Methods based on convolution neural networks, particle swarm optimization, genetic algorithm and D4 (Disjoint Diffusion Deep Face Detection) are used for detection. Authors also discuss the challenges such as generalization of models across the new data, the continuous evolution of adversarial perturbations that leads to consistent attacks, and the scalability issues for the real time deep fake. Concluding that models can significantly improve the accuracy, robustness and generalization.
Authors - Krisha Zalaria, Jaitej Singh, Priyanka Patel Abstract - The ubiquitous use of mobile phones in modern society has sparked increasing concern in environments where their usage is restricted, such as hospitals, schools, religious sites, and hazardous zones. Mobile phones, although integral to daily life, pose risks such as privacy breaches, interference with sensitive equipment, and even serious safety hazards. In response, this paper investigates the efficacy of various state-of-the-art object detection models for real-time mobile phone detection in restricted areas. We benchmarked YOLOv8, YOLOv9, EfficientDet, Faster R-CNN, and Mask R-CNN to identify optimal solutions balancing speed, accuracy, and adaptability. This study introduces a two-class detection framework to distinguish between individuals texting or talking on the phone, catering to differing levels of restriction. Evaluations using a customized, diverse dataset reveal YOLOv8 and YOLOv9 as superior, achieving high precision and recall, thus positioning these models as effective solutions for scalable, real-time surveillance systems in sensitive environments. Our research contributes significant insights into mobile phone detection, paving the way for enhanced safety and privacy in restricted zones.
Authors - Aniket Gupta, Chris Dsouza, Sarah Pradhan, Amiya Kumar Tripathy, Phiroj Shaikh Abstract - This paper is based on the development of voice chatbots and their configuration to make sure that e-commerce websites are in compliance with all the customer care requirements. The authors talk about the introduction of natural language processing in an e-commerce company and provide a review of recent developments in that area. The research specifically focuses on natural language processing techniques, steps involved in developing a chatbot, problems encountered during design, and functions and benefits of voice-based chatbot in e-commerce. This A study emphasizes chatbots as tools to support customer service systems. Keywords: Machine Learning, Natural Language Processing, Data Analysis, Customer support system, and CHATBOT.
Authors - Siddharth Lalwani, Abhiram Joshi, Atharva Jagdale, M.V.Munot, R. C. Jaiswal Abstract - Empathetic response generation is a rapidly evolving field focused on developing AI systems capable of recognizing, understanding, and responding to human emotions in a meaningful way. This paper investigates the integration of the Big Five OCEAN personality model with generative AI to generate emotionally relevant, personalized responses tailored to individual users' personality traits. The Big Five model categorizes individuals into five core personality dimensions—Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. By combining this model with advanced generative AI techniques, the system can deliver empathetic responses aligned with users' emotional states and personality profiles. Through the use of various machine learning algorithms, the study demonstrates that incorporating personality traits significantly improves the quality, accuracy, and emotional resonance of AI-generated responses, leading to more effective human-AI interactions.
Authors - Shaga Anoosha, B Seetharamulu Abstract - Brain tumor segmentation is a critical task in medical imaging, essential for accurate diagnosis and treatment planning. Recent advancements in federated learning (FL) and deep learning (DL) offer promising solutions to the challenges posed by traditional centralized learning methods, particularly regarding data privacy and security. This review paper delves into state-of-the-art approaches that FL and DL to enhance brain tumor segmentation. Each institution trains a deep learning model, typically a Convolutional Neural Network (CNN) or a specialized architectures like U-Net on its local dataset. U-Net, particularly effective for image segmentation tasks, consists of an encoder that extracts hierarchical features from MRI scans and a decoder that reconstructs the segmented output, creating a segmentation map outlining tumor boundaries. Instead of sharing raw MRI scans, federated learning allows each institution to share model updates with a central server. The central server aggregates the updates from all participating institutions to create a global model using Federated Averaging, which averages the weights of the local models. The updated global model is then sent back to each institution, which continues training on their local data using this improved model. This iterative process ensures high accuracy, robustness, and privacy preservation, making it a promising approach for collaborative brain tumor detection and segmentation. By combining the strengths of federated learning and deep learning, these state-of-the-art methodologies provide a powerful solution to the challenges posed by traditional centralized models. This integration not only improves segmentation performance but also ensures that sensitive patient data remains secure. As advancements in this field progress, the collaborative use of these state-of-the-art techniques is poised to significantly enhance diagnostic accuracy and improve patient outcomes in medical imaging.
Authors - Adarsh Singh Jadon, Rohit Agrawal, Aditya A. Shastri Abstract - This study investigated the efficacy of various deep learning models in performing sentiment analysis on code-mixed Hinglish text, a hybrid language widely used in digital communication. Hinglish presents unique challenges due to its informal nature, frequent code-switching, and complex linguistic structure. This research leverages datasets from the SemEval-2020 Task 9 competition and employs models such as RNN (LSTM), BERT-LSTM, CNN, and a proposed Hybrid LSTM-GRU with 1D-CNN model. Combining the strengths of LSTM and GRU units along with 1D-CNN, demonstrated superior performance with an accuracy of 93.21%, precision of 93.57%, and recall of 93.02%, along with Sensitivity & Specificity of 93.62% and 93.24% respectively. It also achieved F1 Score of 93.44%. We also evaluated the model on some other parameters, such as PPV, PNV, RPV, and RNV. This model outperformed existing approaches, including the HF-CSA model from the SemEval-2020 dataset, which achieved an accuracy of 76.18%.
Authors - Chetana Shravage, Shubhangi Vairagar, Priya Metri, Akanksha Madhukar Pawar, Bhagyashri Dhananjay Dhande, Siddhi Vaibhav Firodiya, Tanmay Pramod Kale Abstract - Emotion Recognition has gained significant popularity, driven by its wide range of applications. Emotion recognition methods use various human cues such as use of speech, facial expressions, body postures or body gestures. The methods built for emotion recognition use combinations of different human cues together for better accuracy in results. This paper explores different methods which use different human cues.
Authors - Keesari Abhinav Reddy, Vanaparthi Sai Charan, Md. Sufiyan, Puvula Kiranmai, Madhuri. T, M. Venugopala Chari Abstract - The major challenge for road safety and traffic regulation continues to be categorized traffic offenses that include speeding, running of red lights, improper parking, and distracted driving. Recent innovations in artificial intelligence (AI) and machine learning (ML) have made it possible to develop automated systems that can detect and classify varied traffic violations in detail. This paper analyzes studies that have emerged recently, focusing on advanced technologies, including those such as YOLO-based object detection, OCR, integration with IoT, and real-time monitoring. The paper evaluates datasets, performance metrics, and methodologies covering violations including helmet use, lane changing, and the use of a mobile phone while driving. Significant challenges that have been touched upon in the review include issues of data privacy, high computational requirements, and environmental limitations. Some of the encouraging solution includes use of sophisticated deep learning models, big data analytics, sensor fusion, and edge computing as pathways to enhance scalability and reliability. Future effort will include improvement of real-time systems, reduction of false positives, and addressing socio-technical problems. Using approaches that merge existing advances, this paper has suggested some pathways for using AI-driven systems towards the improvement of road safety and adherence to traffic rules.
Authors - Indushree Shetty, Prerna Agrawal, Savita Gandhi Abstract - The Chronic respiratory diseases, including Chronic Obstructive Pulmonary Disease (COPD), Cystic Fibrosis, Chronic Bronchitis, Interstitial Lung Disease (ILD), Pleural Effusion, Pneumothorax, and Mesothelioma contribute significantly to global mortality and morbidity. The lung diseases in India are influenced by various demographic, environmental, and lifestyle factors like air pollution, high smoking rates, climate change and weather patterns, genetic and hereditary factors, etc. This paper highlights the current scenario of various lung diseases affecting Indian population, highest incident being of COPD to the extent of 89%. The study in this paper surveys the comparison of detection of different lung diseases using machine learning in an Indian Scenario with respect to different parameters like diseases predicted, dataset used, source of dataset, findings, limitations, future score, methods used and accuracy. Based on the comparative study, this paper also highlights various research gaps for future scope in an Indian Scenario. By prioritizing the solutions to the identified research gaps, medical practitioners would be able to handle better India's high respiratory disease burden, increasing the likelihood of more dependable and inclusive healthcare solutions.
Authors - Siddhi Mulewar, Abhijay Patil, Gauri Patil, Nikhil Chame, Smita Kulkarni Abstract - E-commerce has completely transformed traditional retail by lowering operating expenses and enabling worldwide access. Online shopping experiences have been further changed by the integration of artificial intelligence (AI) and machine learning (ML), especially with the advent of Fashion Recommendation Methods (FRM) that employ deep learning techniques. This research introduces a unique FRM that uses a single image input to provide tailored fashion suggestions based on user preferences, improving the quality of the shopping experience. Collaborative filtering (CF) is preferred method in this research work, which encourages users to explore a wider range of content and become more engaged. In this research work ResNet50 pre-trained neural networks proposed to extract information from photos, enabling precise and customized fashion recommendations. Comparative studies show that ResNet50 performs better than other CNN models, leading to increased personalization and accuracy. In the highly competitive world of e-commerce, this study emphasizes the potential of AI-driven suggestions to improve the online shopping experience, stimulate user engagement, and foster loyal consumers. VITON is a Virtual Try-On Network that uses images instead of 3D data to overlay clothes on a person’s image. It creates and refines photo-realistic images with natural clothing deformation using a coarse-to-fine strategy.
Authors - Manasa S Desai, Nirmala M B, Veeresh Kumar Y M, Varsha G C, Vinnet Gokhale, Sushma E Roa Abstract - Electronic Know Your Customer (e-KYC) system is essential for banking and identity providers to verify customer identities efficiently. With the widespread adoption of cloud computing, due to its resource efficiency and high accessibility, many sectors have implemented their e-KYC systems on the cloud. This shift, however, raises significant concerns about the security and privacy of e-KYC documents stored in the cloud. Blockchain technology, a recent innovation, offers potential solutions to enhance various application domains, including digital identity verification. This project proposes a Blockchain-based e-KYC system to address these concerns. This system provides a secure, efficient, and reliable method for identity authentication, which is particularly beneficial in sectors such as banking, tele communications, and government services. By utilizing a distributed ledger to store and verify customer data, the proposed e-KYC framework ensures data integrity and minimizes fraudulent activities. In this framework, customer data is stored on a distributed ledger and encrypted to enhance security. This encryption safeguards sensitive personal information from unauthorized access and cyber threats. This project combines the Ethereum blockchain with Zero-Knowledge Proof (ZKP) technology to provide strong digital identity verification, maintain data integrity, and reduce fraud. The decentralized nature of proposed e-KYC system not only boosts security but also reduces reliance on central authorities, thereby accelerating the verification process and lowering operational costs. This approach offers arobust solution for secure digital identity verification.
Authors - Chilakala Sudhamani, Akula Spoorthi, B. Srilatha Abstract - In today’s world, women face numerous safety challenges, including harassment and molestation. In this paper, we proposed a self-defense stun gun as an effective and efficient solution for women’s safety. This portable device contains a high-voltage generator, GSM and GPS module, panic and taser button and an Arduino Uno with Atmega328 AVR microcontroller. When the device is activated in a dangerous situation, it immediately sends an SMS with the user’s location and distress signal to pre-selected contacts. It also generates a 1000kV electric shock to temporarily immobilize an attacker, allowing the user to escape or seek help. This device aims to enhance the safety and security of women in urgent need or dangerous circumstances for proactive measures against gender-based violence.
Authors - Riddhi Sonawane, Ganesh Bhutkar, Swarup Vishwas, Vivek Badade, Akshay Shingote Abstract - Traditional persona classification methods rely on static, time consuming techniques like surveys and interviews. To address this limitation, we propose F², a novel approach that leverages facial recognition and digital footprint analysis for dynamic persona classification. By integrating real-time data from various digital platforms, F² creates more accurate and up-to-date user profiles. Our system prioritizes user privacy and adheres to relevant data protection regulations. Through robust facial recognition and advanced machine learning algorithms, F² effectively categorizes users into distinct personas, enabling tailored experiences and personalized interactions. This innovative approach has the potential to revolutionize user modeling and enhance digital experiences across diverse domains.
Authors - Aafiya Anjum Abdul Rafique, Martin H Mollay, Shailesh Gahane, Deepak S. Sharma, Pankajkumar Anawade Abstract - There have been remarkable adoptions and uses of Information Technology (IT); therefore, there has been a significant surge in energy consumption and carbon emissions in recent times. While most industries are increasingly relying on digital technologies, IT operations are also increasing their impact on the environment, thereby making green IT a vital necessity. Green IT is an all-encompassing method of managing the environmental footprint of IT through the reduction of energy consumption, electronics waste, and optimum resource efficiency. This paper discusses, from a critical perspective, the role of Green IT in reducing the carbon footprint of IT operations through sustainable technologies and practices. Beyond this, it also discusses challenges and potential solutions for a more green IT landscape in the data center, cloud computing, virtualization, energy-efficient hardware, and new sustainable development practices in software. To sum up, this paper focuses attention on some of the critical factors for driving the adoption of sustainable IT solutions: policy, education, and cross-sector collaboration.
Authors - Dheekshitha Bazar, Gajelli Sai Susmitha, Shreshta Myana, Ramu Kuchipudi, Ramakrishna Kolikipogu, P. Ramesh Babu, K. Gangadhara Rao Abstract - Strategic planning, grid management, and lessening the financial burden on Telangana’s power sector all depend on accurate demand forecasts for electricity. Currently, forecasting methods rely primarily on traditional approaches, but these models often fall short in capturing complex demand patterns at multiple time intervals, especially in dynamic sectors like agriculture. Existing forecasting methods, focused mainly on traditional approaches, often fall short in capturing complex demand patterns across multiple time scales, particularly in sectors like agriculture. This study introduces a comprehensive multi-scale forecasting model for Telangana’s electricity consumption over the next five years, targeting yearly, monthly, weekly, and daily intervals, with a focus on peak load forecasting. Time series techniques such as ARIMA, Prophet, Weighted Moving Average (WMA), and Error Trend Seasonality (ETS) are leveraged to capture seasonality, trends, and short-term fluctuations in demand, providing actionable insights for the Telangana SLDC. Methods for machine learning such as Long Short-Term Memory (LSTM) networks and Gradient Boosting Machines (GBM), are integrated to capture complex temporal patterns and improve predictive accuracy. This study offers a scalable framework for electricity demand forecasting, adaptable to other regions and utilities, advancing methodologies in the power sector. The suggested approach uses metrics to assess the model’s performance such as Root Mean Square Error, Mean Absolute Error (MAE), Both Mean Absolute Percentage Error (MAPE) and RMSE are used to choose the most precise model for every period.
Authors - Hemal S, Sohana R, M Shahina Parveen, Tarun Pradeep Kumar Abstract - Childhood fever poses a significant health concern in India, necessitating timely intervention and effective healthcare strategies. However, predicting fever prevalence accurately remains a challenge due to the diverse healthcare landscape and maternal-child health indicators. This research aims to develop a systematic methodology for predicting childhood fever prevalence based on maternal and child healthcare indicators in India. Leveraging machine learning algorithms, particularly Support Vector Regression (SVR), the study seeks to provide an effective tool for early detection and intervention in infant fever cases. Using data from the "India - Annual Health Survey (AHS) 2012-13" dataset, specific maternal and child healthcare indicators relevant to childhood fever prevalence are identified. These indicators encompass ante-natal care, delivery care, immunization, breastfeeding, and supplementation practices. Various regression algorithms, including SVR, are trained and evaluated to accurately predict childhood fever prevalence. Experimental results demonstrate that SVR outperforms other regression algorithms, showcasing its effectiveness in capturing non-linear relationships and handling outliers. This study offers a structured framework for early detection and intervention in childhood fever cases, leveraging machine learning algorithms and maternal-child health indicators. By accurately predicting fever prevalence, healthcare practitioners can implement timely interventions, ultimately improving healthcare outcomes for infants in India.
Authors - Shivam Kumar Singh, Sindhu Chandra Sekharan, Aishwarya Mondal, Nitin Nagar, Shruti Shreya, Yuting Zhu Abstract - This work presents a comprehensive IoT-based smart assistant device aimed at providing essential navigation and safety support for physically challenged individuals, especially those with visual impairments. The device is equipped with advanced functionalities, including GPS tracking for real-time location monitoring, MobileNet-based object and face recognition, OCR capabilities for reading printed text, and ultrasonic sensors for detecting obstacles, which trigger an alarm to alert the user. Its design prioritizes energy efficiency, allowing it to run effectively on low power while offering reliable real-time processing. By combining multiple assistive features into a single, cost-effective, and portable device, this solution sets itself apart from traditional options that often focus on one functionality or rely on expensive hardware. The modular and scalable architecture not only makes it an affordable and practical solution but also allows for easy customization and potential wireless enhancements. This flexibility opens up possibilities for broader applications in fields like assistive healthcare, autonomous navigation, and consumer electronics, making it a pioneering tool in inclusive technology that enhances mobility, security, and overall independence for its users.
Authors - Sakshi Limkar, Chhandavi Gowardhan, Devyani Dahake, Sneha Naik, Arti Vasant Bang Abstract - Chatbots powered by artificial intelligence (AI) are becoming more and more inventive tools in the field of mental health treatment. They provide scalable, affordable, and easily accessible support for people struggling with stress, anxiety, depression, and other mental health conditions. These conversational bots provide real-time therapeutic interventions, such as promoting emotional well-being by mimicking human interaction through Natural Language Processing (NLP) and Machine Learning (ML) techniques. We have therefore developed a chatbot for mental health assistance named CalmConnect. It is designed to assist users in identifying and addressing mental health concerns.
Authors - Akshar Thakor, Tanya Khunteta, Kaushal Shah, Hargeet Kaur Abstract - It is essential to control access to data. Since the beginning of digital communication, scholars have been at work figuring out ways to prevent eavesdropping on data. They have been successful in doing so through cryptography techniques such as RSA, and AES. Modern Quantum computing is shown to be able to break them. Thus, research and development in this field have been rapid in the last decade. It is better to take precautions in its infancy and develop a future-proof cryptography technique. The following article describes the issues with contemporary ciphers and how they are vulnerable against quantum computers, then goes on to suggest lattice-based cryptography as a strong contender for the solution to this vulnerability by providing its benefits and properties synergizing with certain domains in digital technology. It explains why it is a strong contender by providing examples of its performance in IoT devices that are prevalent today and are only going to increase as this era progresses. By providing a comprehensive overview of developments in this realm, it presents to, new researchers in this field, the importance of Lattice-Based Cryptography, by suggesting why Lattice-Based Cryptography should be the focus of the field of cryptography in the future.
Authors - Nitin Pandit, Sandeep Chaware, Adit Bagati, Yashraj Shegokar, Omkar Jadhav, Om Nikam Abstract - DOS attacks or denial of service have become common among hackers who use them as a way to gain reputation and respect in the cyber underground. A denial-of-service attack essentially means denying legitimate and user network services to a target network or server. Its main purpose is to attack so that legitimate users are temporarily unable to use the services on the network. In other words, we can define a DOS attack as an attack that clogs the target’s memory, making legitimate users unable to help. Or, you send packets that the target cannot process, causing the target to fail, reboot, or deny service to legitimate users. We develop an online DOS protection software that can protect web servers.
Authors - Vedant Patil, Bhargavi Bhende, Omkar Jadhav, Gitanjali Shinde, Kavita Moholkar Abstract - The increasing realism of AI-generated faces, driven by advancements in Generative Adversarial Networks (GANs) like StyleGAN and ProGAN, poses significant challenges in security, identity verification, and digital forensics. Current detection methods, primarily relying on Convolutional Neural Networks (CNNs), struggle to identify subtle artifacts in high-quality synthetic imagery. This paper proposes a hybrid model combining Vision Transformers (ViT) and XceptionNet in a soft-voting ensemble framework. ViT captures global spatial patterns, while XceptionNet excels in detecting localized texture inconsistencies. The ensemble achieves 92.3% accuracy, 92.5% precision, and an F1-score of 0.922 on a dataset of 188,800 real and AI-generated faces. Extensive experiments demonstrate the model’s robustness against diverse deepfake architectures, including those with minimal artifacts. This approach offers a state-of-the-art solution for differentiating real and AI-generated faces, with significant implications for fraud prevention, content moderation, and digital forensics.
Authors - Rupali Ramdas Shevale, Monika Sharad Deshmukh Abstract - For efficient real-time decision-making in a variety of domains, including cybersecurity, finance, and the Internet of Things, accurate and trustworthy event categorization is crucial. By maximizing feature integration, this study explores how incorporating Redpanda, a real-time data streaming platform, into predictive algorithms might improve event categorization. Continuous, high-throughput data processing is made possible by Redpanda's low-latency, fault-tolerant architecture, which enables the real-time extraction of a variety of accurate attributes. Predictive models may use Redpanda's capability to access current, augmented feature sets, which will greatly increase classification accuracy and dependability. The integration process is thoroughly examined in the research, along with its effects on feature variety, model accuracy, and system robustness. The benefits of real-time data streaming in predictive analytics are demonstrated by empirical results, which indicate a significant boost in event categorization performance. By improving feature extraction and enhancing the dependability of predictive systems in dynamic contexts, the results establish Redpanda as a scalable and robust solution.
Authors - Jayashri D.Palkar, Anuradha S. Deshpande Abstract - The crop protection plays vital roles in the food supply and depends on how healthy the crops are, which influences the agricultural production; any adverse condition on crops will be leading to economic loss. Grapes find much use, being important and widely cultivated crops primarily in the Mediterranean regions that control an outgoing market of over 189 billion United States dollars. They are grown for consumption as fresh fruits, as well as in various processed forms such as drinks and sweets. These would be grapes, which, unlike many other plants, thrive and develop despite sickness, thus their control mechanisms must also function well. At the same time, many instances of diagnosis of these infections being wrong can lead to inadequate treatments for the known diseases, inducing even more generalized losses amounting from 5-80% on the crop under inspection. Current computer-based solutions may not be precise enough, leading to high running costs, operational difficulties, and image quality issues due to distortions. The body of literature based on different algorithms for the detection and classification of grape crop diseases remains vast and continues to grow rapidly with the newly emerging algorithms. It presents the overview of different disease-detection algorithms for optimizing grape disease detection, thereby aiding farmers in choosing the appropriate algorithm based on particular diseases and weather condition. This study presents a systematic review of various methods implemented in literature and provides a framework for use of AI-ML for effective detection of disease.
Authors - Shilpa M Katikar, Vikas B Maral, Nagaraju Bogiri, Vilas D Ghonge, Pawan S Malik, Suyash B Karkhele Abstract - Effective forecasting and modeling in food demand supply chains are critical to minimizing waste, reducing costs, and ensuring product availability. This paper explores a comprehensive approach to forecasting food demand by leveraging regression-based models for analysis. We investigate how various machine learning regressors can predict food demand more accurately by examining key supply chain factors such as seasonal trends, price fluctuations, and consumer behavior. The study implements and compares multiple regressors to assess their performance in predicting demand. Metrics Evaluation is done by predicting various models which are Ensemble Learning Models and Neural Network Models to calculate the model’s accuracy. By doing prediction, we identified that Gradient Boosting and XGBoost have overall good accuracy in forecasting and it has provided optimized solutions in the supply food chain. This research mainly focuses on using the best modeling techniques which will help the end users to make proper decisions and bring efficiency in food demand management.
Authors - S. B. Hema Anjali, Manikanta Sai Sumeeth, Sushama Rani Dutta Abstract - This study makes use of a machine learning system that predicts health insurance costs, a relevant issue given the increasing need for such estimates in a post-COVID-19 world. Using the Medical Cost Personal Dataset available at Kaggle offering 1,338 entries, we applied various ensemble models, notably XGBoost, Gradient Boosting Machine (GBM), Random Forest, and Support Vector Machines (SVM). Among our results, XGBoost gives out the best accuracy of the estimates, but the implementation of this technique was expensive. Random Forest was non-intrusive and went on to be of high efficacy. We also discussed how the big data paradigm was implemented using Spark as a means to enhance performance in working on large datasets. As a whole, this work positions XGBoost the ban for the cost of health insurance prediction claiming that there exists scope for improvement by deploying ML methods in decision making in healthcare processes.
Authors - Shrikant Bhopale, Tahseen Mulla, Madhav Salunkhe, Sagarkumar Dange, Sagar Patil, Rohit Raut Abstract - Cardio-Vascular Disease (CVD) continues to be a prominent issue in worldwide health, emphasizing the crucial importance of accurate forecasting and timely prevention. Machine learning (ML) has become a vital tool in the quest to improve CVD diagnosis. The present study aims to conduct a comparative analysis of various machine learning (ML) algorithms in terms of their performance, which includes Naïve Bayes, Logistic Regression, Random Forest, Decision Tree, Artificial Neural Network, Support Vector Machine and XGBoost, in the prediction of CVD. Our results reveal that XGBoost outshines other models, achieving outstanding accuracy, precision, recall, and F-measure. Its exceptional ability to balance precision and recall makes it an excellent choice for the early identification of CVD. This study makes a valuable addition to the expanding field of study on CVD prediction. It underscores the significance of employing advanced ML algorithms, that have the possibility to significantly influence public health outcomes.
Authors - Yathin Reddy Duvuru, Seshank Mahadev, Saranya P Abstract - In this paper, we implement a deep learning model for photovoltaic (PV) power forecasting using Global Horizontal Irradiance (GHI) values which are the major determiner of photovoltaic cell power output. We use a multilayer Long Short-Term Memory (LSTM) model combined with explainable AI (XAI) techniques, aimed at improving the interpretability of predictions across various forecasting horizons. The model utilizes global horizontal irradiance (GHI) data, which undergoes thorough pre-processing, including cleaning and downsampling to ensure data quality and computational efficiency. The LSTM model is designed with multiple layers to capture temporal dependencies and nonlinearities, which are crucial for accurately forecasting PV power under variable environmental conditions. To evaluate model performance, multiple error metrics such as R², MAE, RMSE, and MAPE are utilized. In addition, a benchmark model is built as a reference to compare against the LSTM-based model, providing a baseline for assessing performance improvements. The use of XAI further enables the interpretation of the LSTM model’s predictions, providing an understanding of feature importance and model behavior. We use the SHAP library to perform XAI analysis by calculating Shapley Values. We demonstrate how the SHAP library can be used on 3D LSTM data. Furthermore, the SHAP graphs provide a sense of the importance of each feature’s role in the prediction.
Authors - A.Kousar Nikhath, J.Ananya Reddy, P.Aishwarya, S.Maadhurya Sri, P.Gowthami Abstract - Sign language is the medium between the people who can hear and speak and those who cannot. This project is set to be used in the development of technologies that are beneficial for the lives of individuals with disabilities. The project studies in-depth the use of computer vision and deep learning. The accurate and the regional language translation began with the gestures of the sign languages as the input information, and finally the software produced the accurate translation in the regional language. I am inspired by the prospect of using Artificial Intelligence technology in developing hereditary transmission from a worldwide venue and health diagnosis in a timely manner. Convolutional Neural Network (CNN) is employed to pick up characteristics from hand movement that belongs to the sign language. These attributes are used in the training set as features, themselves in the classification of the gesture, and the process is the learning of this model for the recognition of gestures in real-time. Further, there is an inclusion of computer vision for preprocessing and the sake of accuracy prediction of the recognition process. The functionality of the sign language recognition system is assessed by using a variety of experiments, including accuracy and speed. In general, the developed Sign Language Recognition System with integration of deep learning and computer vision techniques facilitates the precise and quick recognition of sign language gestures. Integration with a translator in addition to this not only makes it multi-language support but also guarantees the correct translation into regional languages.
Authors - Hritesh Kumar Shanty, Padirolu Moses, Tulasiram Nimmagadda, Samson Anosh Babu Parisapogu Abstract - In today’s digital world, combining image editing with secure NFT trading is essential. Imagify addresses this need by offering a unified platform with advanced artificial intelligence tools for image enhancement, recoloring, restoration, and object removal, empowering users to customize images to their preferences. Imagify also simplifies the NFT creation process, allowing users to seamlessly transform their edited images into NFTs that can be bought and sold on a blockchain-secured marketplace. This ensures transparent and secure transactions, providing peace of mind for both creators and buyers. With a flexible, credit-based system, users pay only for the features they choose, making it a cost-effective option. By merging intuitive image editing with a streamlined NFT marketplace, Imagify offers an accessible, user-friendly platform where creators and collectors can engage in digital image trading confidently. This integration creates an efficient and transparent process, supporting both casual creators and seasoned collectors seeking a secure, comprehensive solution for managing and trading digital images.
Authors - R V S S Surya Abhishek, T Sridevi Abstract - This paper provides an overview of the current state of AI-based approaches in virtual fitness coaching, focusing on posture estimation and exercise tracking along with real-time feedback. Advances in pose estimation models, including OpenPose, MediaPipe, and AlphaPose, are boosting personalized exercise correction and injury prevention within the sphere of fitness applications. Current literature varies from 2D to 3D pose estimation that includes action recognition and deep learning framework for specific inputs toward movement analysis and user engagement. There is still much room for improvement in current models, with regards to adaptation to individual needs and environments, such as the real-time accuracy that often has not been matched by the personal feedback and robustness of exercise variations. It discusses the approaches currently in use, their applications, and challenges, and by looking at the topic, this paper insinuates the improvement in the adaptability and customization of AI fitness solutions to perfectly emulate human trainers.
Authors - Mrudul Dixit, Rajiya Landage, Prachi Raut Abstract - The paper presents a comprehensive comparison of Speech-to-Text (STT) and Text-to-Speech (TTS) models, two foundational technologies in the field of natural language processing and human-computer interaction. The paper examines the evolution of these models, focusing on state-of-the-art approaches such as Whisper Automatic Speech Recognition (ASR), DeepSpeech, and Wav2vec, Kaldi, SpeechBrain for STT, and Tacotron, WaveNet, gTTS and FastSpeech for TTS. Through an analysis of architectures, performance metrics, and applications, the paper highlights the strengths and limitations of each model, particularly in domains requiring high accuracy, multilingual support, and real-time processing. The paper also explores the challenges faced by STT and TTS systems, including handling diverse languages, background noise, and generating natural-sounding speech. There are recent advances in end-to-end models, transfer learning, and multimodal approaches that are pushing the boundaries of both STT and TTS technologies. By providing a detailed comparison and identifying future research directions, this review aims to guide researchers and practitioners in selecting and developing speech models for various applications, particularly in enhancing accessibility for specially-abled individuals.
Authors - A.Kousar Nikhath, Aanchal Jain, Ananya D, Ramana Teja Abstract - The project focuses on creating an advanced system for visual speech recognition by performing lipreading at the sentence level. Traditional approaches, which were limited to word-level recognition, often lacked sufficient contextual understanding and real-world usability. This work aims to overcome those limitations by utilizing cutting-edge deep learning models, such as CNNs, RNNs, and hybrid architectures, to effectively process visual inputs and generate coherent speech predictions. The system's development follows a systematic approach, beginning with a review of existing solutions and their shortcomings. The proposed framework captures both temporal and spatial dynamics of lip movements using specialized neural networks, significantly enhancing the accuracy of sentence-level predictions. Extensive testing on diverse datasets validates the system’s efficiency, scalability, and practical applications. This study underscores the critical role of robust feature extraction, sequential data modeling, and hierarchical processing in achieving effective sentence-level lipreading. The results demonstrate notable improvements in performance metrics. Additionally, the project outlines future advancements, including optimizing the system for real-time processing and resource-constrained environments, paving the way for practical implementation in multiple fields.
Authors - Randeep Singh Klair, Gurkunwar Singh, Ritik Verma, Satvik Rawal, Rajan Kakkar, Agamnoor Singh Vasir, Nilimp Rathore Abstract - The most accurate way to measure galaxy redshifts is using spectroscopy, but it takes a lot of computer power and telescope time. Despite their speed and scalability, photometric techniques are less precise. Thanks to large astronomical datasets, machine learning has become a potent technique for increasing cosmology research’s scalability and accuracy. On datasets such as the Sloan Digital Sky Survey, algorithms such as k-Nearest Neighbors, Random Forests, Support Vector Machines, Gradient Boosting, and Neural Networks are assessed using metrics like R-squared, Mean Absolute Error, and Root Mean Square Error. Ensemble approaches provide reliable accuracy, whereas neural networks are excellent at capturing non-linear correlations. Improvements in feature selection, hyperparameter tuning, and interpretability are essential to improving machine learning applications for photometric redshift estimation and providing deeper insights into cosmic structure and development.
Authors - Sudha K L, Navya Holla K, Kavita Guddad Abstract - The antenna is a vital component of the Magnetic Resonance Imaging (MRI) machine which receives the radio frequency signals emitted by the protons in the body after the RF pulse is turned off. Specialized high frequency antennas can improve the quality, clarity, and resolution of the resulting MRI images. This paper deals with the design of Bow Tie antenna for X-Band in the frequency range 8–12 GHz, used in ultra-high field MRI systems. Using the Ansys HFSS tool, the antenna is designed and simulated and analysed. The fabricated antenna with the design specifications is tested in anechoic chamber for its working. Reflection coefficient at 10.5GHz is found to be around -14 dB for simulated antenna and -12 dB for fabricated antenna, which is satisfactory for practical application. Differences between the measured and simulated values were seen in results which are caused by cable loss in the measuring apparatus.
Authors - Bharati B Pannyagol, S.L Deshpande, Rohit Kaliwal, Bharati Chilad Abstract - The Internet of Things has revolutionized markets by connecting previously isolated devices, but this integration raises security risks from malicious nodes that can corrupt data or disrupt operations. This evaluation of Federated Learning's possible application as a decentralized node identification technique highlights its advantages over standard machine learning approaches. Internet of Thing devices may collaborate on model training while protecting sensitive data and reducing network use. Federated Learning and Blockchain interactions creates a robust framework addressing critical IoT challenges like data privacy, security, and trust. Blockchain enhances this system by providing a decentralized, tamper-resistant ledger that ensures data integrity and transparency. Automated processes, including model validation and incentive distribution, are facilitated by smart contracts. While this integrated approach improves data protection and scalability, challenges such as computational demands and consensus delays remain. The survey discusses practical applications, challenges, and future research directions for combining Federated Learning and Blockchain in IoT systems.
Authors - Bhadouriya Khushi Mukeshsingh, Rajput Adityasingh Shashikantsingh, Patel Swayam Vinodkumar, Ashish P. Patel, Nirav D. Mehta, Anwarul M. Haque Abstract - As digital infrastructure becomes more interconnected, effective cyber security has never been more important. This article explodes how advances in power electronics technology can support and improve cyber security frameworks. Energy management strategies, control systems, and semiconductor technologies can be used to increase the systems resilience to potential vulnerabilities that serve as possible entry points for cyber attackers. The research discussed in this article seeks to demonstrate that optimized distribution systems with adaptive control techniques can improve the stability and reliability of critical infrastructure, even in the face of cyber threats. This article discusses the inter-relationship between energy management and cyber security, showing the reader how power electronics can be important in developing a holistic security strategy. It describes a proposed approach to integrating power electronics into cyber security to create an adaptive, robust defence mechanism. This study provides valuable insights into the design of systems that are not only efficient but also fortified against evolving cyber threats, contributing to the broader understanding of how technology convergence can enhance overall infrastructure security.
Authors - Madhura Shankarpure, Dipti D. Patil Abstract - This paper presents a robust framework for YOLO (You Only Look Once) algorithm- based orange detection and localization in photos and videos is presented. The system combines contour-based bounding box localization with deep learning-based item recognition for increased accuracy. Transfer learning was used to refine a pre-trained YOLOv10 model on a Fruit 360 dataset. Data augmentation techniques such as random rotations, brightness changes, and scaling were applied to improve the model's resilience. Bounding boxes are created around identified oranges with a confidence threshold greater than 0.5 as part of the real-time video processing methodology. The model performed well on a balanced test dataset, achieving 95% accuracy, 92% precision, and 90% recall. These findings show how well YOLO works when combined with conventional computer vision methods for real-world uses like automated fruit sorting, fruit harvesting, and real-time market monitoring. The processed video output confirms the system's suitability for real-world situations.
Authors - Nomusa Vumisa, Hendrik Pretorius, Marie Hattingh Abstract - Persuasive features in video games play an important role in encour-aging continuous indulgence and behavior change. Research has been conducted to investigate the different motives, game design elements and features that en-hance the gaming experience. However, there is a gap in understanding how per-suasive features in video games have an impact on individuals, resulting in be-havioral changes. An understanding of the role each feature plays in a video game is crucial for the successful creation and design of a video game aimed to “per-suade” players to change their behavior. This paper presents a systematic review that covers 30 publications and is aimed at investigating the persuasive features in video games and further providing a classification of those features. The results of the study provide a guide to the main theories on behavior change and a clas-sification of the identified persuasive features. Additionally, this study provides a reference for video game designers and developers to utilize when undertaking persuasive projects.
Authors - Sahil Thakur, Saloni Mahadule, Palash Singh Chandel, Sudhanshu Maurya, Firdous Sadaf M. Ismail, Rachit Garg Abstract - As CCTV technology has continued to mature quickly, important and fundamental questions about secure, scalable and transparent storage are posed. Most traditional stored-concentrated models can have challenges on the data’s integrity since other unauthorized users may easily manipulate or delete the video data. This paper explores the design of a decentralized CCTV surveillance system and with motion detection and preprocessing and cloud computing technology. Focused on motion detection, video frames are recorded and compressed with the help of OpenCV and then stored on AWS S3 for further instant access and in AWS Glacier as final storage. Each of the defined operations—upload, deletion, and modification—of the stored video frames is logged transparently on the Ethereum block chain. AWS provides scale and security to the cloud, and block chain provides for the possibility of non-tamperable records. This architecture does not only secure the video data from other people’s violation as well as prevent themselves from being permanently erased but also manage massive video data. The findings presented here show that integration of AWS cloud services with block chain could provide a highly secure, scalable, and transparent solution for today’s CCTV systems.
Authors - Shailaja B. Jadhav, D. V. Kodavade, Suhasini S. Goilkar Abstract - Data centric applications are increasing worldwide, inspiring data scientists to devise more sophisticated methods capable of modelling highly dynamic, extremely speedy data. There are existing approaches which adopt concept learning, dynamism, combining different approaches and heterogeneous classifiers. But, very few of them consider real time data generated through live data savvy applications. This necessitates Streaming data analytics as emerging area of research traditional data mining is not sufficient to achieve desired efficacy. This research aims to focus on streaming data classification particularly flight stream data and presents a comprehensive design framework of multi-layered ensemble built through pool of classifiers selected with prequential evaluation. The model is experimented with various known platforms of streaming data analysis like scikit multiflow, MOA etc. through systematic experimental work. Also, considering the volume of streaming data the experiments have also utilised GPU environments and Google TensorFlow wherever necessary. This Research addresses data streaming analytics majorly, as it needs more attention from research community. There is still scarcity of established benchmarks and standardized frameworks. Major observations, evaluation of design finds that the designed model is able to capture the dynamic nature and improves the classification accuracy as compared with that of the available traditional ensemble models.
Authors - S. P. Vibhute, S. C. Patil, S. A. Bhisikar Abstract - The relentless advancement of wireless communication technologies has ushered in the era of 6G networks, necessitating innovative strategies to enhance their energy efficiency without compromising performance. This study addresses the critical need for sustainable and efficient 6G networks, particularly in the context of growing environmental concerns and escalating energy demands. Existing models, while foundational, often fall short in optimizing energy consumption, grappling with issues such as high latency, increased complexity, substantial costs, and limited scalability. To bridge this gap, our work systematically reviews and analyses various models, including Sleep Scheduling and Intelligent Routing, among others, to augment the energy efficiency of 6G networks. The review process is comprehensive and multidimensional, comparing these models across key performance metrics such as delay, complexity, cost, energy efficiency, and scalability. By employing a meticulous and structured approach, this study elucidates the strengths and limitations of each model, providing a holistic understanding of their applicability in real-world scenarios. The implications of this work are far-reaching, offering invaluable insights for stakeholders in the wireless communication domain. It equips practitioners and researchers with empirical evidence to identify and implement the most optimal models, thereby significantly enhancing the efficiency and sustainability of 6G networks. The findings from this study are poised to contribute substantially to the development of more robust, energy-efficient, and scalable wireless communication systems, aligning with the global drive towards sustainable technological advancements.
Authors - The Tung Than, Thi Phuong Nhi Le, Dong Thanh Vo, Minh Son Nguyen Abstract - Object tracking is crucial in computer vision, particularly in robotics, but Visual Object Tracking (VOT) faces significant challenges, with occlusion being the most critical. Occlusion disrupts tracking accuracy and poses difficulties when integrating VOT algorithms into embedded robotic systems due to computational and real-time constraints. To address this, we propose a robust method tailored for resource-limited systems, combining the Kernelized Correlation Filter (KCF) and Kalman Filter (KF). By leveraging the Average Peak-to-Correlation Energy (APCE) index, our method detects occlusion, dynamically adjusts the model’s learning rate, and improves performance under challenging conditions. Experimental results on the OTB-100 benchmark highlight our tracker’s effectiveness in handling occlusion, achieving a success rate of 0.602. This demonstrates the method’s robustness under challenging conditions while maintaining real-time processing at 30 FPS (Frame per Second) on Jetson Nano, making it an ideal solution for embedded robotic systems.
Authors - Pulipaka Hrishitha, Hima Atluri, Kovvur Ram Mohan Rao Abstract - In this study, we evaluate two distinct chatbot models integrated into a comprehensive healthcare platform, with a focus on addressing medical and mental health inquiries. The chatbot, driven by LLAMA 2, equipped with an inbuilt Retrieval-Augmented Generation (RAG) mechanism, specializes in retrieving and generating precise responses to medical queries and is tailored to offer personalized and empathetic support for mental health concerns. Through meticulous analysis, we assess the effectiveness of these chatbots against a spectrum of functional and non-functional requirements, encompassing usability, security, scalability, accuracy, and empathy. Our investigation delves into LLAMA 2's performance across four distinct scenarios related to mental health inquiries. These scenarios involve variations in fine-tuning and the provision of custom prompts to the chatbot. We scrutinize LLAMA 2's responsiveness in both finetuned and non-fine-tuned states, as well as with and without custom prompts, aiming to discern the impact of these optimization strategies on the chatbot's capacity to deliver empathetic and supportive responses. Our findings provide valuable insights into the nuanced intricacies of LLAMA 2's role in mental health support within AI-driven healthcare solutions, offering implications for further development and refinement in this critical domain.
Authors - V.S.N. Murthy, S. Sai Nikitha, T. Nithya Sri, P. Sushma, V. Kavya Harshitha, M. Hema Lalitha Abstract - Users today often have to navigate multiple websites to find information that matches their preferred level of understanding and format, whether it's text, audio, or video. This fragmented approach can be time-consuming and frustrating, especially when seeking information suited to specific needs. The challenge is to find a streamlined solution that provides comprehensive, and customizable content in one application eliminating the need for users to jump from one site to another. Our project resolves this issue by developing a unified web-based application that allows users to input a topic and select their preferred content format and level of understanding. The platform uses LLM to generate detailed information and seamlessly convert it into the chosen format, whether it be text, audio, or video. This integrated approach ensures that users receive the information they need in their desired format, all in one convenient location. By simplifying the process, our platform provides a more efficient and user-friendly way to access information suited to their preferences, making it easier for users to learn.
Authors - Lakshya Khanna, Shriniwas Mahajan, Varun Kadu, Sudhanshu Maurya, Firdous Sadaf M. Ismail, Rachit Garg Abstract - Sales forecasting assumes a significant part in essential navigation and asset allotment for organizations across different businesses. Realizing the patterns can change and help in the plan of market procedure, particularly now, when the Electronic Vehicles (EV) market is at its pinnacle. In this paper, we investigate the utilization of measurable models and some high-level AI procedures, specifically Random Forest and Long Short-Term Memory (LSTM) models, for anticipating sales information patterns. The review plans to assess the exhibition and dependability of these models in estimating sales data, utilizing genuine world datasets spreading over several years. Execution assessment of the models is led utilizing measurements like Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared. Also, stability analysis is performed to evaluate the unwavering quality of each model in catching and foreseeing exact patterns. The discoveries of the exploration feature the viability of the measurable models and ML models in anticipating sales data patterns. The two kinds of models show promising execution, with the LSTM model areas of strength for displaying in catching transient conditions and long-haul designs in the sales data. In any case, contrasts in execution and strength between the models are noticed, giving important experiences to choosing the most appropriate determining approach in view of explicit business prerequisites.
Authors - Ravi Kumar Suggala, Penumala Syamya, Pokuri Venkata Naga Rohitha, Nunna Reshma Sri Hanu, Vuyyuri Gnana Prasuna, Vegesana Naga Sai Pujitha Abstract - Dental cavity identification using advanced image processing and machine learning techniques, especially through X-rays, plays a crucial role in early diagnosis and treatment planning. Traditional detection systems often suffer from high error rates and low accuracy. To address these challenges, a sophisticated model based on Riemannian Residual Neural Networks with Improved Sooty Tern Optimization (RR2Net-ImSTOpt) is proposed. The model uses the DENTEX dataset for analysis, incorporating noise reduction and image enhancement using Guided Box Filtering (GBF). Feature extraction is performed using the Inception Vis-Transformer, followed by optimization of RR2Net's weight parameters via the Improved Sooty Tern Optimization Algorithm. This approach achieves impressive results with a recall of 99.8% and an accuracy rate of 99.9%, surpassing current methods in accuracy and reducing false positives. RR2Net-ImSTOpt’s capability to handle large medical datasets makes it an ideal solution for clinical dental cavity detection, enhancing diagnostic efficiency and precision.
Authors - Aisha Karigar, Mohammed Qadir Ternikar, Harsh Nesari, Vanashree N, Prema T. Akkasaligar Abstract - Problematic Internet Use (PIU) is a growing concern, especially among adolescents, with significant impacts on mental and physical health. This study aims to predict the severity of PIU, measured by the Severity Impairment Index (SII), using a combination of physical activity, demographic, and behavioral data. Machine learning models, including XGBoost, CatBoost, TabNet, and LightGBM, were employed to classify participants into SII categories: none, mild, moderate, and severe. Data were sourced from the Healthy Brain Network (HBN) dataset, which includes accelerometer data, internet usage, fitness assessments, and physiological measures from over 3,000 participants aged 5 to 22 years. Key feature engineering steps included creating interaction terms (e.g., BMI × Age) and applying Autoencoders for dimensionality reduction on the high-dimensional actigraphy data. The results indicated that CatBoost performed best in predicting minority SII categories, handling imbalanced data effectively. XGBoost and LightGBM demonstrated stable performance, while TabNet provided interpretability but lower overall predictive power. Evaluation metrics, particularly Quadratic Weighted Kappa (QWK), were used to assess model performance, with QWK offering insights into the ordinal nature of misclassifications. This study highlights the value of combining physical activity and behavioral data in predicting PIU severity. The findings underscore the potential of machine learning in identifying individuals at risk for severe PIU and suggest avenues for future interventions to reduce the negative impacts of excessive internet use.
Authors - K Sai Geethanjali, Nidhi Umashankar, Rajesh I S, Jagannathan K, Manjunath Sargur Krishnamurthy, Maithri C Abstract - This survey provides a comprehensive review of the methods used for lung cancer detection through thoracic CT images, focusing on various image processing techniques and machine learning algorithms. Initially, the paper discusses the anatomy and functionality of the lungs within the respiratory system. The review examines image processing methods such as cleft detection, rib and bone identification, and segmentation of the lung, bronchi, and pulmonary veins. A detailed literature review covers both basic image enhancement techniques and advanced machine learning methods, including Random Forests (RF), Decision Trees (DT), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and Gradient Boosting. The review highlights the necessity for reliable validation techniques, explores alternative technologies, and addresses ethical issues associated with the use of patient data. The findings aim to assist researchers and practitioners in developing more accurate and efficient diagnostic tools for lung cancer detection by providing a concise review, thereby helping to save time and focus efforts on the most promising advancements.
Authors - Bijeesh TV, Bejoy BJ, Krishna Sreekumar, T Punitha Reddy Abstract - Integrating artificial intelligence (AI) and advanced imaging technologies in medical diagnostics is revolutionizing brain tumor recurrence prediction. This study aims to develop a precise prognosis model following Gamma Knife radiation therapy by utilizing state-of-the-art architectures such as EfficientNetV2 and Vision Transformers (ViTs), alongside transfer learning. The research identifies complex patterns and features in brain tumor images by leveraging pre-trained models on large-scale image datasets, enabling more accurate and reliable recurrence predictions. EfficientNetV2 and Vision Transformers (ViTs) produced prediction accuracy of 98.1% and 94.85% respectively. The study’s comprehensive development lifecycle includes dataset collection, preparation, model training, and evaluation, with rigorous testing to ensure performance and clinical relevance. Successful implementation of the proposed model will significantly enhance clinical decision-making, providing critical insights into patient prognosis and treatment strategies. By improving the prediction of tumor recurrence, this research advances neuro-oncology, enhances patient outcomes, and personalizes treatment plans. This approach enhances training efficiency and generalization to unseen data, ultimately increasing the clinical utility of the predictive model in real-world healthcare settings.
Authors - Melwin Lewis, Gaurav Mishra, Sahil Singh, Sana Shaikh Abstract - This paper focuses on the development of a 2D Fighting Game, using Simple DirectMedia Layer 2 (“SDL2”), Good Game Peace Out (“GGPO”) and the Godot Game Engine. This project was made with the help of the Godot Engine and the prototype test was implemented in the C++ Language with the intent to showcase the GGPO library for the implementation of Rollback Networking in Fighting Games, a technology which makes seamless online-play possible without input delay.
Authors - Anudeep Arora, Neha Tomer, Vibha Soni, Neha Arora, Anil Kumar Gupta, Lida Mariam George, Ranjeeta Kaur, Prashant Vats Abstract - Improving patient outcomes, maximizing operational efficiency, and guiding strategic decision-making all depend on the capacity to analyze and interpret data effectively in the quickly changing healthcare sector. Finding and analyzing outliers is a major difficulty in healthcare analytics as it can have a big influence on the accuracy and dependability of data-driven conclusions. The significance of business outlier analysis in healthcare analytics is examined in this article, along with its methods, uses, and consequences for payers, providers, and legislators. Healthcare companies may improve their analytical skills, which will improve patient care by improving forecast accuracy and resource allocation. This can be achieved by detecting and resolving outliers.
Authors - Aswini N, Kavitha D Abstract - Obstacle detection is vital for safe navigation in autonomous driving; however, adverse weather conditions like fog, rain, low light, and snow can compromise image quality and reduce detection accuracy. This paper presents a pipeline to enhance image quality under extreme conditions using traditional image processing techniques, followed by obstacle detection with the You Only Look Once (YOLO) deep learning model. Initially, image quality is improved using Contrast Limited Adaptive Histogram Equalization (CLAHE) followed by bilateral filtering to enhance visibility and preserve edge details. The enhanced images are then processed by pre-trained YOLO v7 model for obstacle detection. This approach highlights the effectiveness of integrating traditional enhancement techniques with deep learning for robust obstacle detection, even under adverse weather, offering a promising solution for enhancing autonomous vehicle reliability.
Authors - Hetansh Shah, Himangi Agrawal, Dhaval Shah Abstract - The paper outlines the design, implementation, and evaluation the SHA-256 cryptographic hash function on an FPGA platform, focusing on its use in Bitcoin mining. SHA-256 is a key part of the Bitcoin system, generating unique hash values from data to keep it secure and intact. The goal was to create a fast and low resource utilized, hardware-based version of SHA-256 using VHDL and implement it on the Zed- Board FPGA development platform. The main focus was on the VHDL implementation, making it modular and pipelined to improve speed and efficiency regarding resource utilization. The Zed-Board features the Xilinx Zynq-7000 SoC has been considered for hardware implementation. The design also included message buffering, preprocessing, and a pipeline for hash computation, allowing the system to handle incoming data in real time while producing hash outputs quickly. The algorithm’s functionality was verified using simulation tools in Xilinx Vivado, and the hardware implementation results were compared to previous works. It is clearly depicted the proposed method utilizes fewer resources as compared to the previous works while maintaining a throughput 27% greater than the software solution. The hardware design significantly outperforms software as well as SW/HW (HLS) versions in speed and energy use. The total on-chip power utilized was 12.898 W.
Authors - Janwale Asaram Pandurang, Minal Dutta, Savita Mohurle Abstract - Black Friday shopping event is one of the most awaited events worldwide now a day, it offers huge discounts and promotions of various products categories. For sellers, it’s important to know the customer purchasing behaviors during this period to predict sales, manage inventory and planning for marketing strategies. This research paper will focus on developing a machine learning model that will predict customer expenses capacity based on previous data from Black Friday, by considering factors such as demographics, product types and previous purchases. After collecting and processing a different dataset, exploratory data analysis was conducted to find important trends. Different machine learning models, like linear-regression, K-nearest-Neighbors (KNN) Regression, Decision-Tree-Regression and Random-Forest-Regression, were applied and tested. The Regression Forest Model with R2 value of 0.81, was found with strong predictive accuracy among those models. This study focuses on machine learning models which will help sellers to improve their productivity and will increase revenue.
Authors - Srikaanth Chockalingam, Saummya B. Gaikwad, Lokesh P. Shengolkar, Dhanbir S. Sethi Abstract - This paper presents an innovative microcontroller-based system designed to convert text files into Braille script, making Braille content more accessible for visually impaired users. The system leverages an ARM-based microcontroller and servo motors to enable real-time, mechanical translation of text into tactile Braille characters. To facilitate ease of use and to allow offline operation, an SD card is used as the primary storage medium for text files, enabling users to load and convert documents without requiring an internet connection or additional devices. This design emphasizes affordability, scalability, and usability, with the primary aim of making Braille conversion technology more accessible to educational institutions, libraries, and individuals, particularly in resource-limited settings. By reducing dependency on costly, proprietary Braille technology, this system can improve access to information and literacy among visually impaired communities, especially in developing countries where Braille materials are often scarce or prohibitively expensive. The paper thoroughly explores the system’s hardware and software components, detailing the architecture and function of each element within the overall design. A focus on energy efficiency is highlighted to extend the device’s operational time, and efforts to minimize manufacturing costs ensure this solution remains within a low-cost budget. These design choices make this Braille converter a sustainable option for broad deployment and adoption. Further development aims to expand the device's functionality by integrating wireless connectivity for text input, allowing users to access a greater range of content through online sources. Additionally, future iterations could support a larger tactile display, accommodating more Braille cells simultaneously, which would improve the reading experience for users and enhance the system’s application in educational environments.
Authors - Olukayode Oki, Abayomi Agbeyangi, Jose Lukose Abstract - Subsistence farming is an essential means of livelihood in numerous areas of Sub-Saharan Africa, with a significant segment of the population depending on it for food security. However, animal welfare in these agricultural systems encounters persistent challenges due to resource constraints and insufficient infrastructure. In recent years, technological integration has been seen as a viable answer to these difficulties by enhancing livestock monitoring, healthcare, and overall farm management. This study investigates the effects of technological integration on enhancing animal well-being, with an emphasis on a case study from Nxarhuni Village in the Eastern Cape province of South Africa. The study employs a random sampling method of 63 subsistence farmers to investigate the intricacies of technology adoption in rural areas, highlighting the necessity for informed strategies and sustainable agricultural practices. Both descriptive and regression analyses were employed to highlight the trends, relationships, and significant predictors of technology adoption. The descriptive analysis reveals that 56.6% of respondents had a positive perception of technology, even though challenges like animal health concerns, environmental conditions, and financial constraints persist. Regression analysis results indicate that socioeconomic status (coef = 1.4468, p = 0.059) and gender (coef = -1.1786, p = 0.062) are key predictors of technology adoption. The study recommends the need for specialised educational programs, improvement in infrastructure, and community engagement to support sustainable technology use and enhance animal care practices.
Authors - Dinesh Rajput, Prajwal Nimbone, Siddhesh Kasat, Mousami Munot, Rupesh Jaiswal Abstract - We introduce a system based on neural networks that combines real-time avatar functionality with TTS synthesis. The which system can produce speech in the voices of various talkers, including ones that were not seen during training. To generate a speaker embedding from a brief reference voice sample, the system makes use of a unique encoder that was trained using a large volume of voice data. Using this speaker voice, the algorithm converts text into a mel-spectrogram graph, and a vocoder turns it into an audio waveform. Concurrently, the produced speech is synced with a three-dimensional avatar that produces equivalent lip motions in real time. By using this method, the encoder's learned speaker variability is transferred to the TTS job, enabling it to mimic genuine conversation in the voices of unseen speakers. On a web interface, precise lip syncing of speech with facial movements is ensured via the integration of the avatar system. We also demonstrate that The system's ability to adapt to novel voices is markedly improved by training the encoder on a diverse speaker dataset. In addition, The capacity of the model to generate unique voices that are distinct from those heard during training and retain smooth synchronization with the avatar's visual output is demonstrated by the use of random speaker embeddings, which further showcases the model's capacity to produce high-caliber, interactive voice cloning experiences.
Authors - Ravin Kumar Abstract - This paper introduces the Adaptive Base Representation (ABR) Theorem and proposes a novel number system that offers a structured alternative to the binary number system for digital computers. The ABR number system enables each decimal number to be represented uniquely and using the same number of bits, n, as the binary encoding. Theoretical foundations and mathematical formulations demonstrate that ABR can encode the same integer range as binary, validating its potential as a viable alternative. Additionally, the ABR number system is compatible with existing data compression algorithms like Huffman coding and arithmetic coding, as well as error detection and correction mechanisms such as Hamming codes. We further explore practical applications, including digital steganography, to illustrate the utility of ABR in information theory and digital encoding, suggesting that the ABR number system could inspire new approaches in digital data representation and computational design.
Authors - Aarya Pendharkar, Tanmay Pampatwar, Mrunal Zombade, Ashwini Bankar Abstract - This study offers an effective approach for forecasting changes in stock prices using a binary classification model that makes use of sentiment analysis, technical indicators, and historical stock data. The model forecasts whether a stock will gain or lose the following day, rather than predicting actual stock prices. Technical indicators including moving averages, the Relative Strength Index (RSI), and Bollinger Bands are among the input elements, along with historical price data (open, close, high, low, and volume). Market news and social media data are subjected to sentiment analysis, which produces sentiment ratings (positive, neutral, or negative) in order to identify general patterns in market sentiment. When combined with technical indicators, these mood scores provide additional context for stock movements. The model uses machine learning techniques like XGBoost, SVC, Logistic Regression, and Random Forest, and it outputs a confidence score and a binary forecast. Performance indicators like accuracy, precision, recall, and F1 score are used to assess the model's efficacy. Back testing is also done to evaluate the robustness and performance of the past. The suggested model offers a comprehensive perspective of stock movements by integrating technical and sentimental aspects, producing better prediction skills than conventional models that only use past price data.
Authors - Divyashree H.B., Shirshendu Roy, Supraja Eduru, Dev Sharma, Prathamesh M.Naik Abstract - In today's tech scenario maximum farmers are practicing unconventional farming which needs hard work, in detail if say it is physical practicing. Especially the day-to-day work if talk about that watering the crop manually without measuring the temperature or having the knowledge of soil moisture in the field. As this is practiced from generation to generation, instead of any gain they are losing manpower, water loss which leads to low production and lower the income of farmer. The development of smart agriculture which is built, gives the surety about the soil's water level and fertility outcome by using several sensors. The sensors which are included is temperature sensors, soil moisture sensors and humidity sensors. The coordinated work with these sensors integrated with IoT and raspberry pi will make it convenient and limits the excessive work of the farmers. The integrated sensor will be placed on the water tank and interconnected with pump source, will give alert notification to the farmer phone about the need of water supply. Most the problems are related to electricity is there this issue can be resolved by connecting the sensors with power source and integrating it with cloud so that every controls of the farm will be in the fingertips of farmers. Similarly for soil moisture sensors in case of water requirement by the soil will be directly reach to users phone. So they can perform irrigation. Cattles responsibility is there, farmers owns livestock in the time of grazing, it may lost or distracted from the pathway. Collar tracker with map support will be beneficial at that time. Livestock abnormal behaviors can be detected, there feeding and water tank refilling can be done by just one click. Cows milk thickness health issue and certain things can be managed. Not only limited to cow but for other livestocks. Climate and weather conditions will be directly updated on the applications. Data analytics support for managing expenses. Graph guidance for the soil moisture, temperature and irrigation support. Water tank percentage filled, air composition whether drop irrigation or sprinkler irrigation needed, temperature, humidity cattles live location on custom based maps will be displayed on the dashboard. Application usage guidance and query support will be there for smooth use of application.
Authors - Ankit Patne, Hritika Phapale, Kaushik Aduri, Hemantkumar B Mali Abstract - Cloud-based Learning Management Systems (LMS) are secure online platforms that enable L&D professionals to upload their resources and build a comprehensive suite of learning materials. This paper presents an overview of the cloud LMS technologies landscape and examines architecture, scalability solutions, and security perspectives on deploying these tools. We take a look at how these platforms are also incorporating machine learning into their personalization of learning experiences. If you investigate some of the case studies on platforms like Coursera, you will get a sense of practical ways to implement and maintain performance improvements. The present paper, by reviewing current literature reviews major benefits of cloud technologies in improving educational outcomes, which include reducing cost, better scalability, and enhanced security. Such a study contributing to the evolving knowledge base of cloud-based education is shedding new light into the possibility of how cloud LMS can revolutionize IT security education delivery.
Authors - R.Mehala, K.Mahesh Abstract - The Content Preserve for 3D Video Stabilization using Warping Techniques for making a hand-held video camera captured using a guided camera motion. This technique enables the simulation of 3D camera movements by modifying the video look it was captured from adjacent views. Its algorithms successfully reproduce dynamic scenes from a single source video by focusing solely on perceptual plausibility rather than perfect reconstruction. The method that modifies a hand-held video camera's output to make it look as though it captured using a directed camera motion. This technique enables the simulation of 3D camera motions by modifying the video to look as though it was captured from adjacent views. It is possible to automatically select a particular wanted camera path. The warp calculated the content maintains the video frame while adhering to sparse deletions suggested by the restored 3D structure. This method works well as seen by the experiments stabilizing difficult movies with dynamic sceneries.
Authors - K.Sarvani, Dinesh, Bijith Narayanan, Aayush Rai Abstract - Digital finance has become a buzzword in every financial service to identify any country's solvency position and competitive environment. This study emphasizes the performance of the banking sector with respective to macro-economic variables to assess the solvency and profitability position of commercial banks in India. Two macroeconomic variables namely gross domestic product, and inflation were considered to identify the performance of nonperforming assets of the public sector banks. There are twelve public sector banks in India as of 2013-24 as per the RBI database. All the public sector banks were considered for the study for ten years. The data was collected from PROWSSIQ for the financial data of public sector banks. Macroeconomic variables were taken from Economic Times data from the published data from web sources. The findings of the study are that non-performing is negatively correlated to inflation and GDP growth rates. The adjusted R Squared value is 61 percent implying that the regressors are perfectly explained that the dependent and independent had a relation. Forecasting the performance of non-performing was done using the SARIMA model. It is found that for all the select banks, non-performing assets are continuously increasing which implies that the recovery of bad debts may be done by the adoption of new fintech apps and it is a positive sign for the performance of the banks in coming years.
Authors - Anudeep Arora, Neha Arora, Neha Tomer, Ranjeeta Kaur, Vibha Soni, Lida Mariam George, Anil Kumar Gupta, Prashant Vats Abstract - Effective human resource management is a major issue for the Business Process Outsourcing (BPO) business, which is marked by a high staff turnover rate and a dynamic operating environment. These issues are frequently not adequately addressed by traditional HR management techniques, which results in inefficiencies and higher expenses. BPO companies may improve employee engagement, optimize staffing levels, and anticipate workforce demands with the use of predictive analytics, which makes it a potent option. The use of predictive analytics for efficient HR utilization in the BPO sector is examined in this article. It explores important technologies, tools, and processes; talks about the advantages and difficulties of implementation; and provides case studies of effective deployments. BPO firms may increase labor productivity, lower attrition, and boost overall company success by utilizing predictive analytics.
Authors - Kavita Patil, Rohit Patil, Vedanti Koyande, Amaya Singh Thakur, Kshitij Kadam, Kavita Moholkar Abstract - This paper evaluates a chatbot system designed for personalized business interactions using advanced Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). The system combines proprietary business data with external databases to improve contextual relevance. A comparative analysis of leading LLMs—Gemini Pro, GPT-4, Claude 2, GPT-3.5, and LLaMA 2—was conducted across benchmarks like MMLU, GSM8K, BigBench Hard, HumanEval, and DROP. Gemini Pro outperformed the others, with scores of 88.9% on MMLU, 86.3% on GSM8K, 78.1% on BigBench Hard, 73.5% on HumanEval, and 79.2% on DROP, showcasing its strength in complex reasoning and long-context retrieval. Fine-tuned with business-specific data, Gemini Pro sets a new standard for high-accuracy, scalable chatbot solutions, ideal for enterprise applications.
Authors - Darshana Naik, Aishwarya Bhagat, Amman Baheti, Atharva Kulkarni, Hitesh Kumar Abstract - This paper examines the potential of AI-powered chatbots to address the growing global need for accessible and effective mental health support. It traces the evolution of chatbots, from rudimentary systems to sophisticated AI-driven platforms, emphasizing advancements in artificial intelligence and natural language processing that enable personalized responses. Driven by the need to overcome barriers of cost, availability, and stigma in mental health care, the paper explores chatbot integration strategies. These include using chatbots for screening and triage, extending therapist reach, bridging care gaps, reaching underserved populations, and leveraging data for personalized interventions. While chatbots show promise in delivering therapeutic support and improving symptoms, they are envisioned as a complement to, rather than a replacement for, traditional therapy. The paper advocates for leveraging AI to enhance the scalability, reach, and personalization of mental health care, ultimately aiming to improve global mental health outcomes. By exploring both the potential and the challenges of AI-powered chatbots, this paper contributes to the ongoing dialogue about the future of mental health care in an increasingly digital world.
Authors - Ashwini Bhosale, Laxmi Patil, Gitanjali Netake, Sakshi Surwase, Rutuja Gade, Prema Sahane Abstract - This paper discusses a project that aims to create a system for translating sign language into spoken words while also recognizing the emotions of the signer. The goal is to make communication easier for Deaf and hard-of-hearing individuals by converting hand gestures into speech and reflecting the signer’s emotional tone in the voice output. This would make conversations feel more natural and expressive, enhancing interactions in both social and work environments. The project uses computer vision and Convolutional Neural Networks (CNNs) to accurately recognize various sign language gestures. To identify emotions, it uses deep learning models like VGG-16 and ResNet, which focus on facial expressions. It also uses Long Short-Term Memory (LSTM) networks to analyze audio input and detect emotional tones in speech. For turning sign language into spoken words, the system employs Text-to-Speech (TTS) technologies like Tacotron 2 and WaveGlow. These tools create natural-sounding speech, and the detected emotions are added to the voice by adjusting tone, pitch, and speed to match the signer’s feelings. With real-time processing and an easy-to-use interface, this system aims to provide quick translation and emotion detection. The expected result is a fully functional system that not only translates sign language into speech but also effectively conveys emotions, making communication more inclusive for Deaf and hard-ofhearing individuals.
Authors - Sujith Kumar Banda, Ramzan Shareef, Swathi Sowmya Bavirthi, Mohammed Arbaz Ahmed Abstract - While meetings help make company decision-making more effective, documenting and distilling the material turns out to be a lot of time-consuming work and may also contain mistakes. The project provides an automated way of transcribing audio recording of meetings into text and applying NLP for perfect creation of useful summaries. As opposed to the existing techniques that resort to either means of human beings or platform-specific ones, our solution is a versatile way that can handle transcripts from a variety of online resources. This is a system that offers both abstractive and extractive summary techniques in the form of developed transformer models, such as BERT, to form logical summaries and TF-IDF and TextRank to focus the most important points in the summary. A wider applicability of Named Entity Recognition (NER) and Part-of-Speech (POS) tagging will allow summarization over key elements, including decisions taken and responsibilities assigned. The approach aims to make the capture of output from the meeting more efficient and reliable by automatically summarizing proceedings in meetings. User input and ROUGE scores will assess how well the system performs and guarantees quality useful summaries to stakeholders.
Authors - Rajeev Sharma, Santanu Sikdar, Govind Murari Upadhyay Abstract - To protect network infrastructure from new vulnerabilities and security dangers caused by the rapid growth of Internet of Things (IoT) devices, robust and adaptable Intrusion Detection Systems (IDS) are necessary. Due to their limited scalability and reactivity to different attack patterns, conventional intrusion detection systems (IDS) struggle to meet the unique demands of Internet of Things (IoT) networks. The novel Intrusion Detection System introduced in this paper is based on deep learning and is tailor-made for Internet of Things (IoT) environments. It employs complex neural network topologies to enhance the accuracy and efficiency of detection. Regarding the massive amount and variety of data generated by IoT devices, our suggested method improves performance without compromising detection accuracy by combining feature selection and dimensionality reduction strategy. Standard IoT network datasets were used for training and validation, with several assaults implemented to ensure comprehensive threat coverage and practical applicability. The results of the experiments show that the proposed system outperforms the state-of-the-art machine learning-based intrusion detection systems in detection accuracy, false positive rates, and scalability in contexts with limited resources for the Internet of Things.
Authors - Rutuj Barudwale, Vijeyandra Shahu Abstract - This article focuses on attempting artificial intelligence in stock price forecasting. Common stock market predictions and their prices can be assessed using dual primary analytical models known as technical and fundamental analysis. I employed a technical analysis of price trends predicting price movements using regression machine learning (ML). For instance, predicting how the price of a particular stock will close at the end of today based on historical price trends. In contrast to this approach of technical analysis, fundamental analysis can be applied to supervised machine learning algorithms to assist with identifying how news and social network users appear to be for or against certain entities. In the technical analysis, the historical price trends are retrieved from Yahoo, and in the fundamental analysis, the stock market tweets are analyzed to assess how the public feels about the stock prices. The findings portray an average performance; therefore, given the present environment of - technology, it is rather optimistic to presume that technology will ever beat the stock market consistently.
Authors - Satvik Taviti, Srreyasri Kurlagunda, Nandikanti Sri Gayatri, R M Krishna Sureddi, Raman Dugyala Abstract - Electronic Health Records (EHR) are considered to be amongst the most crucial elements for exchanging data in healthcare services. Thus, security for these records is the keystone of patient privacy and easy cooperation between the service providers. This review looks at four primary approaches to EHR security and predictive management: Blockchain, Attribute-Based Encryption (ABE), Deep Learning, and Access Control Models. Blockchain ensures data integrity, transparency, and traceability but scalability issues, high transaction costs, and interoperability challenges prevent its widespread adoption. ABE is appropriate for fine-grained access control in data sharing under the patient-centred approach but cumbersome and resource-intensive for managing encryption across the large healthcare network. Deep learning helps predictive analytics, personalized medicine, but with high computational demands that affect its real-time application in the clinical environment. While in terms of data confidentiality protection, models such as Role-Based or Attribute-Based Access Control may ensure proper restriction of authorized access, they might not suffice for dynamic, multi-provider health environments. Comparing the techniques will outline their relative merits, weaknesses, and security considerations, thus helping to understand how safe yet scalable systems for EHR storage could be built.
Authors - Edidiong Akpabio, Supriya Narad Abstract - This review aims to understand the integration of two emerging technologies: artificial intelligence and the Internet of Things. IoT is defined as the capability of implementing connections between regular items and industrial apparatuses that can liaise in real-time, exchanging and analyzing data. AI is an ideal companion to IoT in the sense that it brings decision-making into the equation and boosts the effectiveness and functions of IoT systems. This paper aims to review the use of AI and IoT in various fields, namely, smart cities, health, farming, and transport. In smart cities, IoT and AI applications are also applied to enhance traffic, energy consumption systems, and urban design. It has changed the way that the healthcare industry operates through better methods of patient monitoring, performance analysis, and telehealth. In agriculture, IoT sensors help monitor the effectiveness of crop management and the use of AI-based automation. It also covers the implementation of AI and IoT in autonomous vehicles, particularly the use of sensors for data processing, decision-making, and real-time data communication. However, the use of AI and IoT has some limitations, such as data limitations, security and privacy, and environmental impact. Indeed, the paper dwells upon these issues and provides the outlook for further research regarding edge AI, IoT sustainability, and the further evolution of the connections. With technological progress still in the process of evolving AI and IoT, future advancements hold more potential in terms of creating better connected, efficient, and sustainable solutions, not to mention the fact that AI is capable of solving existing challenges.
Authors - Rydhm Beri, Parul Sachdeva Abstract - The advent of IoT technology has significantly transformed the industrial sector, paving the way for the emergence of Food Industry 4.0. This research explores the integration of edge–cloud computing and IoT to create a smart framework tailored for the food industry. Central to this framework is the appli-cation of a Bayesian belief network (BBN) on an edge–cloud platform, enabling data-driven insights into food quality. The framework assesses data to calculate the Probability of Food Quality (PFQ) and utilizes the Food Quality Analysis Measure (FQAM) to evaluate food outlets. A bi-level decision-tree model further enhances the evaluation process by providing an in-depth analysis of food quality metrics. To address concerns around data security, blockchain technology is implemented, ensuring the protection of food-related information. The model is rigorously tested on a comprehensive dataset encompassing 43,520 instances from four restaurants. Simulation results highlight its high performance, achieving a temporal delay of 96.43 seconds, and the system demonstrates an accuracy of 98.93%, showcasing its robustness in real-world applications.
Authors - Rydhm Beri, Parul Sachdeva Abstract - Plant diseases present a serious threat to all forms of life. Early detection is vital which allows farmers to take prompt action, improving both their response and productivity. Our research centers on five common rice leaf diseases—bacterial leaf blight, leaf blast, brown spot, leaf scald, and narrow brown spot—along with a category for healthy leaves. Additionally, we examine two types of betel leaves: healthy and unhealthy. This study propose an innovative deep ensemble model that combines the EfficientNetV2L, InceptionResNetV2, and Xception architectures. This model addresses issues of underfitting and performance by utilizing advanced techniques including data augmentation, Global Average Pooling, preprocessing, Dropout, L2 regularization, PReLU activation, Batch Normalization, and multiple Dense layers. This robust approach surpasses existing models by managing both underfitting and overfitting, while delivering superior performance.
Authors - T. Sridevi, Chidhrapu Harini, Kurella Sai Veena Abstract - Sign Language is the primary means of communication among 1.8 million deaf people across India, and although Indian Sign Language (ISL) translation to technology-based effective solutions is still very limited, tremendous effort has so far been made in global research in sign language recognition. Nevertheless, the challenge persists in transcoding of text from ISL. This project will fill the vacuum by developing a deep-learning-based model capable of generating subtitles for ISL videos. With a pre-trained Convolutional Neural Network (CNN) for spatial feature extraction and a Recurrent Neural Network (RNN) for encoding the temporal pattern, the model learns on the Indian Sign Language Videos dataset. Designed in a manner to achieve high-accuracy captioning of ISL for reliable communication with the Indian deaf community. This will provide access to means of communication for millions of ISL users, but at the same time offers a critical communication tool meant to facilitate improvement in circles of education, social life, and professional circles in India.
Authors - Anudeep Arora, Vibha Soni, Lida Mariam George, Anil Kumar Gupta, Ranjeeta Kaur, Neha Arora, Neha Tomer, Prashant Vats Abstract - In the field of financial analytics, stock market prediction continues to be one of the most difficult and sought-after objectives. A key component of stock price modeling and forecasting is time series analysis, a statistical technique that examines sequences of data points gathered at successive times. A thorough review of time series analytic techniques for stock market prediction is given in this article. These techniques include machine learning and deep learning, as well as more sophisticated approaches like GARCH and ARIMA. It addresses the drawbacks and advantages of these methods, looks at the difficulties in putting them into practice, and identifies new developments in time series forecasting. Investors and analysts may improve their ability to anticipate the future and make better judgments in the ever-changing stock market environment by being aware of these techniques and how they are used.
Authors - M R Shreyaank, Dhanush Karthikeya A J, Dhanush Rajan S, Ashwini Bhat Abstract - This research attempts to investigate the potential healing effects of Vedic chants and music on the human brain through an in-depth analysis of EEG signals. The Vedic chants are known for their inherent calming and meditative attributes and are believed to impart positive influences on the human mind and body. The study employs a simulative model to analyse EEG signals during exposure to Vedic chants. Recorded EEG signals from MDD (major depressive disorder) subjects are subjected to preprocessing and feature extraction processes involving frequency-domain analysis and power spectral density. The study compares the extracted features between conditions of Vedic chant exposure and controlled settings and shows that there is significant increase in alpha and beta powers after listening to the specified chants. Rejuvinating and Calming chants showed the best positive impact.
Authors - Suresh V Reddy, Sanjay Bhargava Abstract - Cybercrime on social media platforms such as Facebook and Twitter has emerged as a significant challenge due to the open, interactive nature of these platforms. Various machine learning (ML) and deep learning (DL) techniques have been deployed to detect different forms of cybercrime, including phishing, spamming, hate speech, and identity theft. This paper provides a comparative analysis of these approaches, focusing on their application to cybercrime detection on Facebook and Twitter. Through a detailed literature review, we evaluate the strengths and weaknesses of these techniques, considering their performance and scalability. Moreover, the ethical challenges and the need for privacy-preserving mechanisms are discussed, along with future directions for research.
Authors - Rakesh Babu B, Rajesh V, Syed Inthiyaz, Srinivasa Rao K, Sri Sravan V Abstract - Brain tumours are life-threatening disorders with significant fatality rates. Patients have a higher chance of survival when brain tumours are diagnosed early and treated more effectively. Therefore, for the purpose of better and boost the early identification of brain tumours, computerized segmentation as well as classification techniques are needed. It is possible to safely and promptly detect tumours using brain scans such as computed tomography (CT), magnetic resonance imaging (MRI) and other techniques. Revolutionary changes have occurred in many different disciplines as a result of recent developments in artificial intelligence (AI). AI models are becoming essential tools for interpreting images in bio medical field. Deep learning is one of these that signifies extraordinary capacity to deal with enormous data collection, revolutionizing numerous fields in the biomedical profession. This article evaluates a state-of-the-art AI based segmentation and classification system and discovers major classes for brain tumours. The potent learning capability and effectiveness of AI approaches have been assessed. Convolutional Neural Network (CNN) is one of the AI subfields that has demonstrated remarkable performance in analysing medical imagery. Consequently, the processing of medical imagery, particularly brain MRI images, was the main emphasis of this review paper, which also examined different deep learning model architectures in addition to CNN.
Authors - Shubham Kadam, Chhitij Raj, Pankajkumar Anawade, Deepak sharma, Utkarsha Wanjari, Janhvi Shirbhate, Sharvari Pipare Abstract - Artificial Intelligence (AI) is increasingly being hailed as the key to the future of healthcare supply chain management in countries such as India, where healthcare is a particularly complex setting for an integrated supply chain. This review presents the various Data-driven Artificial Intelligence (AI) technologies such as Machine Learning (ML), Natural Language Processing (NLP), Computer Vision, and Robotic Process Automation (RPA) that help in the automation of essential processes like demand forecasting, inventory management, and cold chain logistics in an efficient and timely manner. AI helps deliver vital supplies on time and minimizes any disruptions of services by utilizing predictive analytics and real-time monitoring. However, high implementation costs, data privacy concerns, the need for integration with legacy systems, and a need for more skilled professionals are barriers to the adoption of AI computing. To extract the maximal potential AI can offer healthcare logistics, the issues above need to be addressed. Upcoming research directions include further development in quantum computing, IoT integration, and collaborative AI platforms to fulfil resilience and sustainability objectives for supply chains. The results underscore the potential of AI to transform health supply chains and provide an opportunity to realize more scalable, responsive, and efficient health services.
Authors - Rajeshree Khande, Sachin Naik, Akshay Tayade, Amar Kale, Kunal Phalke Abstract - The authors propose for the LSTM-XGBoost model for portfolio optimization as well as stock price prediction. The model has incorporated the benefits derived from XGBoost, a gradient-boosting algorithm that enhances the ability of a model to predict structured and improved data, and Long Short-Term Memory (LSTM) networks, which excel at characterizing time-series data based on temporal relationships. The XGBoost model takes advantage of the LSTM model by utilizing the anticipated outputs it makes for improving the precision and overall efficiency of the model while the LSTM model is designed to work with ordered data peculiar to stock markets specifically on patterns and trends over time. In the study authors employ this type of hybrid to determine variables such as volatility and the moving average of historical stock price index of NIFTY50. The authors have obtained total model accuracy of 98.33%. Authors also use the Sharpe ratio to maintain an optimal portfolio because it shows investors the optimal ratio of expected stock returns. This research contributes to enhancing financial forecasting by integrating deep learning and machine learning techniques, ultimately offering the formulation of a new risk avert portfolio as well as stock price prediction.
Authors - Jaiditya Nair, Sunil Kumar Abstract - The increasing demand for AI-driven solutions in development has encouraged people to conduct various research into generating code from natural language prompts. My paper presents a Retrieval-Augmented Generation (RAG) pipeline for code generation, making use of embedding models, contextual retrieval, and advanced language models such as Mistral and CodeLLama. This approach incorporates document indexing and metadata extraction to create context-aware code snippets and at the end of the process, we get a python file with the generated code present in it.
Authors - Nishita Shekhar Bala, Sree Vani Bandi, Stephen R, Ravi Dandu, Balakrishnan C Abstract - These days internet is became an essential part of human life and affects various domains which includes education, business, social interactions, mental health. It pushes the society ahead through increasing innovations, amplifying learning techniques, connecting people across the globe and access to vast resources which makes it a valuable tool in this modern society. But it comes with problems such as internet addiction, sleeping disorders, health complications. This abstract discusses about dual impact of internet uses, focusing on its significant benefits and possible dangers. Hence, there is need to mange use of internet so one can make use of its benefits at the same time reducing the affects which are caused by internet on human life.
Authors - Chandan Raj B R, A. Yasaswi, Deepika K, Uday Bhaskar Reddy, Delina Yadav K, Joshna K Abstract - It is quite difficult to communicate with deaf individuals. This article extends the complexity of Indian Sign Language (ISL) character classification. Sign language is insufficient for the hearing and speaking disabled. Hand gestures of disabled individuals may appear confused to those who have not learnt the language. Communication should be two-way. In this essay, we will discuss how to learn a language through sign language. Images are processed using computer vision processes, including grayscale conversion, dilation, and masking. We employ Convolutional Neural Networks (CNN) to train and recognize images. Our example has an accuracy of approximately 95%. Gestures serve as a nonverbal communication tool in language. People with hearing or speech difficulties frequently utilize them to communicate with others or among themselves. Many loudspeakers are created by various manufacturers around the world. This study demonstrates that many experiments are undertaken each year, with several articles published in journals and conferences, and that research on vision-based gesture recognition is ongoing. Cognitive navigation focuses on three areas: information retrieval, environmental information, and gesture representation. In terms of identity verification, we also evaluated the authentication system's effectiveness. The physical movement of the human hand generates gestures, and gesture recognition contributes to improvements in autonomous vehicle operation. This paper use the convolutional neural network (CNN) classification technique to detect and recognize human motions. This workflow consists of region-of-interest coordination via masking, finger segmentation, normalization of segmented finger pictures, and finger recognition using a CNN classifier. Use the mask to separate the hand portion of the image from the rest of the image. The histogram equalization approach is used to improve the contrast of each pixel in an image. This work uses a variety of scanning techniques to classify fingerprints from hand photographs. The segmented fingers from the hand image are put into the CNN classification algorithm, which separates the image into different groups. This research proposes gesture recognition and recognition methods based on CNN classification, and the technology achieves good performance using cutting-edge methodologies.
Authors - Anudeep Arora, Ranjeeta Kaur, Neha Tomer, Vibha Soni, Neha Arora, Anil Kumar Gupta, Lida Mariam George, Prashant Vats Abstract - The incorporation of data analytics into internal audit operations is a noteworthy progression in augmenting the efficacy and productivity of audits. In this paradigm, strategic analysis refers to using data-driven insights to evaluate risks, expedite audit procedures, and enhance organizational controls. This article examines the use of strategic analysis in data analytics and internal audits, including important techniques, advantages, and difficulties. It talks about how sophisticated data analytics methods, such as machine learning, statistical analysis, and visualization software, can change the way that auditing is done today. In addition, the paper looks at case studies and potential future developments in the subject, giving readers a thorough understanding of the various ways internal auditors might use data analytics to provide audit results that are more precise and useful.
Authors - Artika Singh, Manisha Jailia Abstract - Effective management of infectious disease outbreaks rely heavily on informed decision-making processes. There are many approaches given for decision-making some of them are expert decision-making, creative problem solving, public engagement, and decision-making under deep uncertainty (DMDU) in outbreak management (OM). The integration of these aspects is critical to enhancing the responsiveness and efficiency of public health interventions. This paper discusses the current state of expert decision-making processes, the role of creativity in managing complex situations, the impact and challenges of incorporating public and patient engagement (PPE) in OM. The paper concludes with recommendations for future research and practice to improve outbreak management strategies.
Authors - Prateeksha P Malagi, Priyanka R Patil, Shamshuddin K G, Suneeta V Budihal Abstract - The advent of blockchain technology presents a transformative opportunity for enhancing the integrity and efficiency of voting systems. This paper explores the design and implementation of a blockchain-based voting system aimed at addressing common challenges faced in traditional electoral processes, such as voter fraud, lack of transparency, and low participation rates. By leveraging the decentralized and immutable nature of blockchain, our proposed system ensures secure voter authentication, real-time vote tracking, and tamper-proof record keeping. The study outlines the technical architecture, including smart contracts and cryptographic techniques, while evaluating the system's performance through simulated voting scenarios. Furthermore, we discuss the implications of this technology for promoting democratic engagement and restoring public trust in electoral outcomes. Our findings suggest that a blockchain-based voting system not only enhances security and transparency but also offers a scalable solution to modern electoral challenges.
Authors - Manjusha Pandey, Rajeev Kumar, Satyam Tiwary, Yuvraj Singh, Oindrella Chatterjee, Siddharth Swarup Rautaray Abstract - This paper delves into the complexities of providing equitable access to multimedia content across India's diverse linguistic landscape. It proposes innovative strategies for translating English video content into Indian regional languages, leveraging cutting-edge technologies such as machine translation, speech recognition, and text-to-speech synthesis. The suggested approach involves a systematic four-phase process, encompassing audio separation, text conversion, machine translation, and speech synthesis. [1] By utilizing open-source tools like IBM's Watson supercomputer and the Flite engine from Carnegie Mellon University, the system achieves a commendable 79% accuracy in terms of naturalness and fluency, as evaluated by native speakers. However, challenges persist in handling multi-speaker conversations and accommodating a broader range of Indian languages. Despite these limitations, the research lays a solid foundation for future advancements in the field. By fostering cross-cultural communication and knowledge dissemination, the proposed solution holds the potential to bridge linguistic barriers, empower marginalized communities, and foster an inclusive digital ecosystem in India.
Authors - Shubham Kadam, Chhitij Raj, Pankajkumar Anawade, Deepak sharma, Utkarsha Wanjari Abstract - This paper explores the phenomenon of Artificial Intelligence (AI) transformation in organizational culture evaluation, discussing capabilities, advantages, obstacles and future direction. While traditional means of mining forms like surveys and interviews are often lengthy and flawed due to human biases, AI tools rely on real-time data, natural language processing, and predictive analysis to deliver objective insights instantly. Such applications, including sentiment analysis, behavioural analytics, and cultural diagnostics, allow organizations to mitigate cultural misalignments in advance at the organizational level or within specific teams, idem for the employee's engagement and inclusivity. Nonetheless, ethical issues related to data privacy, security and algorithmic wage discrimination continue to pose significant challenges. The implications of this study highlight the increasing importance of artificial intelligence in enabling organizations to build dynamic, resilient, and agile organizational cultures.
Authors - Sakshi Sharma, Tanisha Verma, Shailesh D. Kamble Abstract - Accurate, timely detection of plant disease is critical to protect crop from being damaged and increase agricultural productivity. Many disease identification methods are labor intensive and only practical with an expert set of trained eyes. A mobile application for real time plant disease detection using CNNs presented in this paper allows farmers to have a simple yet powerful access to a diagnostic tool. CNN was trained on a big collection of plant leaf images to discriminate between diseases using Keras and TensorFlow. The application was built using Flutter for cross platform mobile development, trained model deployed on mobile devices using TensorFlow Lite, which allows offline inference. Users can capture images of affected plant leaves and get immediate diagnostic feedback as to the potential disease involved. Following data preprocessing and model optimization, the application uses a lightweight architecture that achieves high accuracy while meeting requirements for mobile deployment. This research shows integration of AI with mobile technology can provide a scalable, efficient and accessible solution to crop disease detection. The system as proposed is capable of improving crop health management, reducing losses, and working towards global food security.
Authors - Utkarsha Wanjari, Shubham Kadam Abstract - Gamification in HRM through AI is thus a total revolution that can maximize the engagement and productivity of employees. Game-like qualities such as rewards, badges, leaderboards, and challenges incorporated in the HR processes create a captivating environment that motivates and pushes an employee into an achievement culture. AI amplifies the effect of gamification: it enables data-driven insights, personalized experience, and real-time feedback loops. The paper also looks into the psychological underpinnings of gamification intrinsic and extrinsic motivation and their alignment with the organizational goals. It analyzes some of the challenges in incorporating gamification, including ethical considerations, potential overuse, and the balance between entertainment and productivity. It also reflects on some success stories and presents a pathway to implementing gamified AI solutions into the existing HR framework. This is because gamification, combined with AI, will alter the way human resource practice prevails, uplift employee productivity, boost employee satisfaction, and contribute to the long-term success of an organization. The present research study aspires to provide business organizations with the actionability of a very innovative method to remain ahead of their game in the changed wilderness of the workplace.
Authors - Shubham Kadam, Chhitij Raj, Pankajkumar Anawade, Deepak sharma, Utkarsha Wanjari Abstract - The paper investigates Green ICT leadership in e-governance towards carbon footprint mitigation from the digital government. E-governance uses information and communication technology (ICT) to deliver administrative services through enhanced technology in this service chain, thus increasing the efficiency of their services, which is guided by an aim for complete transparency that requires accurate information. However, digitalization is responsible for environmental problems such as carbon emissions produced by data centres, digital infrastructure, and devices. The paper emphasizes the importance of vision-oriented leadership in promoting sustainability through processes of strategic thinking, collaboration and innovation. The Guide presents a series of critical strategies, including energy-efficient data centres, virtualization and cloud computing, sustainable procurement, and citizen engagement to build green practices. Innovative technologies such as AI, IoT, and blockchain are labelled enablers for optimizing energy consumption and increasing transparency.
Authors - Alpa R. Barad, Ankit R. Bhavsar Abstract - Analysis of grass quality is essential to improve cattle health. To improve animals' health and productivity, it is necessary to survey quality food. Grass is a primary and major source of food for every cattle. As a part of vegetation quality of grass is decreasing day by day, and it’s also not possible to survey fresh grass on a daily basis. Proposed research is used to analyze the quality of grass based on its color space. The quality of grass differs over the grass species and weather, and it's become more difficult with a single model to recognize its quality. To solve this problem proposed research uses machine learning based hybrid approach. The proposed research uses Median filter with kmeans clustering. Based on the clustering, the Simulation uses color deflection code to identify threshold values for a given species of grass. Proposed research finds the remarkable performance of three different qualities of grass. Simulation of study uses a Wiener filter and data augmentation to identify the impact of the proposed k-means based hybrid approach for grass quality recognition.
Authors - Aryan Jain, Shrirang Joshi, Vatsal Jain, Dinesh Kumar Saini Abstract - 5G network roll-out is expanding globally, which further shows that low-cost and good modem design remains to be absolutely integral. Scaling here is tough, not to mention the complexity and cost of production involved in traditional hardware-based 5G modems. This analysis explores how advances such as Open Radio Access Networks (Open RAN), Software-Defined Networking (SDN) and Network Function Virtualization (NFV) could reduce the hardware requirements, leading to lower costs for 5G modems. We marvel over the functionalities which we take for granted in a modem, such as digital signal processing and base-band processing, are being virtualized so that it is done on general-purpose hardware rather than on parts custom designed to do these particular tasks. Adopting cloud-native and software-based solutions for these traditional hardware-driven processes can bring huge savings without compromising on performance. In addition, we discuss Dynamic Resource and Change in Network efficiency which are improved by Modem Allocation, Edge Computing, Network Slicing — SDN NFV open day light. This collection of methods is described in a comprehensive article on the application of virtual network technologies to improve 5G modem design, reduce deployment costs, and enable more flexible, scalable, and energy-efficient 5G solutions.
Authors - Mallu Praneeth Reddy, T. A. S. Vardhan, Kura Bhargava Gupta, Nagireddy Deekshitha, Pudari Shrainya Goud, Khalvida Pamarty, Sushama Rani Dutta Abstract - This paper aims to predict the calories burnt by a person using machine learning models built on several regression algorithms like Linear, Random Forest, XGBoost,and CatBoost based on gender, age, height, weight, duration of exercise, body temperature, and heartbeat of the person. In addition, the analysis compares the algorithms based on performance metrics like MAE (Mean Absolute Error), MSE (Mean Square Error), and R2 score and determines the most effective algorithm for calorie prediction.
Authors - Shailender Vats, Prasadu Peddi, Prashant Vats Abstract - Blockchain technology's explosive growth has created previously unheard-of potential in several industries, but it has also revealed fresh security flaws. To improve threat detection and response mechanisms, this paper provides a complete intrusion detection system (IDS) designed especially for distributed blockchain ledger security. It makes use of sophisticated smart contracts. We demonstrate the efficacy of the suggested IDS in detecting possible intrusions while preserving the integrity of the blockchain environment by validating it using simulation-based scenarios. According to the research, combining IDS with blockchain technology and smart contracts greatly improves security and is a viable way to address current cybersecurity issues.
Authors - M Nanda Kumar, Harsh Sharma, Rajan Kakkar, Tushar Naha, Atul, Rishabh Yadav, Naveen Abstract - The demand for autonomous vehicles (AVs) has grown rapidly due to their potential to revolutionize transportation by enhancing safety, efficiency, and convenience while reducing human error, a leading cause of road accidents. AVs leverage advanced technologies like machine learning, LIDAR, GPS, cameras, RADAR, and ultrasonic sensors for precise navigation, obstacle detection, and real-time decision-making. However, their reliability and safety in di-verse environmental conditions remain a significant challenge. Extreme weather events such as heavy rain, snow, fog, ice, hail, and dust storms can impair sensor performance, reducing visibility, traction, and the ability to detect road markings, obstacles, and other vehicles. These conditions degrade the accuracy of critical systems like LIDAR, RADAR, and cameras, raising concerns about AVs’ reliability, particularly in emergencies or unpredictable scenarios. This review paper explores the effects of adverse weather on AVs’ performance, analyzing the limitations of key sensors and assessing various mitigation strategies to enhance their resilience. By identifying technological gaps and emphasizing the need for weather-resilient solutions, the paper aims to guide future research and innovation to improve AVs’ safety and reliability in challenging real-world conditions, ensuring their readiness for broader deployment.
Authors - Sneha Singh, Deepak Kaushal, Bhupinder Preet bedi, Sanjay Taneja, Pawan Kumar Abstract - An ever-increasing number of individuals from all over the world are devoting a significant portion of their time to activities that take place in the digital realm, such as communicating with one another and looking for information. There is no denying the fact that social media platforms, which include Facebook, Twitter, sites like Instagram, and video sharing platforms like YouTube, play a vital role in the day-to-day lives of individuals, thereby altering the way in which people go about their routines. Over the past few years, electronic word-of-mouth communication, often known as eWOM, has seen a significant surge in popularity. Accordingly, the purpose of the study is to gain an understanding of the current situation regarding eWOM and social networks by means of a comprehensive review of the relevant literature. A comprehensive selection of 100 research studies was obtained from Scopus. The findings will offer a new direction to academicians in the future.
Authors - Akash K, Joseph Jilvis J, Felicia Lilian J, Subhashni R Abstract - This paper focuses on a voting system which is on a blockchain technology platform. To address issues that are known to be present in voting, it employs decentralized applications of ethereum known as dApps. Some of the contemporary matters raised include fraud as well as complexity. The proposed dApp is based on the use of smart contracts, as well as two-factor authentication through Metamask. A number of features might be noted. For example, one of the services provided by the system is event coverage such as the elections results. There is also a Voter Analysis Report Feature. This particular feature provides information on demography and the voting behaviour and it best viewed in pie chart. This dApp employs technologies including HTML, CSS, JavaScript & solidity in the process of its development. All in all, it seeks to enhance integrity, accessibility and transparency of the voting system. By doing this, it intends to increase trust and openness in elections more effectively.
Authors - Eshwari Khurd, Tushar Nasery, Rupesh C. Jaiswal Abstract - Data storage and applications have observed a large shift from being stored and used in local drives just a decade ago to being almost entirely cloud dependent today. This change in usage has brought about new challenges to be dealt with. Traditional security solutions were developed keeping in mind the use case for local storage. Techniques like cryptography have evolved to be more adaptive and secure. Yet, time after time, it has been proven that they can be broken. However, this is no longer adequate as the working and use of cloud networks is vastly different than local storage devices. Thus, new solutions need to be developed in order to secure this already established pattern of data consumption.
Authors: Shailesh T. Khandare, Nileshsingh V. Thakur Abstract: Image segmentation is the key and important process in the image analysis. In general, thresholding technique is used for the grey level image segmentation and when it comes to apply for the color images, the RGB color image is separated in three grey level planes and then it is applied on these grey level planes or else the color image is directly converted to grey level image and then it is applied on this converted grey level image. This paper addresses the issue of computation time requirement to carry out these three grey level plane image segmentations through the generation of grey level image without using any inbuilt function of tool or platform. The data fusion approach is proposed which is based on the trichromatic coefficients. A single grey level image is formed from the available IR, IG, and IB grey level planes using the trichromatic coefficients. Obtained results are compared on the basis of bilevel and multi-level thresholds. Otsu bilevel threshold of obtained grey level image differs with the Otsu bilevel threshold of converted grey level image by 11 %. Obtained grey level image by proposed approach is visually near about similar to converted grey level image. Error between the thresholded images of proposed approach and converted image is less. Obtained multi-level threshold values are close with the multi-level threshold of converted image.
Authors - Rajni, Parminder Kaur, Harmandar Kaur Abstract - The current world is run by technology and network connections, which are indispensable parts of day-to-day life. Corporate organizations, the military, and the government have adopted automation, and computers connected to the network are being used for the storage and sharing of vital, highly confidential, and valuable information. Hence, is essential to prevent the attackers from exploiting the vulnerabilities for illegally accessing the crucial data. With increased dependency on the internet owing to the proliferation of technologies such as cloud computing, the Internet of Things (IoT), wireless communication, and social media networks, high security is required in cyberspace. Cybersecurity provides the methods used to protect sensitive information in cyberspace. Disturbed denial of services (DDoS), phishing, man-in-the-middle (MiTM), passwords, SQL injection, Cross-site scripting, malware, and drive-by download are a few types of cyberattacks. Traditional methods such as firewalls, intrusion detection systems, antivirus software, access control lists, etc., are no longer productive in detecting new generation attacks. Therefore, there is an urgent need to design new methods to prevent these sophisticated cyberattacks. This paper explains the main reasons for cyberattacks and reviews the various types of cyberattacks, their vulnerabilities, detection and prevention techniques. To prevent current and future cyberattacks such technologies as machine learning, cloud platforms, big data, and blockchain can play an important role. The solutions provided by these technologies may assist in detecting malware, intrusion detection, spam identification, DNS attack classification, fraud detection, recognizing hidden channels, and distinguishing advanced persistent threats, enhancing the overall defense against sophisticated cyberattacks.
Authors - Shubham Garg, Kanika Monga, Nitin Chaturvedi, S. Gurunarayanan Abstract - Approximate computing has emerged as a promising paradigm for error- tolerant AI/ML applications deployed on energy-constrained edge devices. It has gained significance for edge devices due to its potential to reduce power consumption. In conventional computing systems, implementing computationally intensive machine learning algorithms results in large power consumption. Addressing this challenge, the complexity of hardware computing units can be reduced by optimizing the circuit logic while slightly trading off the computational accuracy. This technique is termed as Approximate computing where the circuit provides close-to-accurate results rather than precise results with significant reduction in power consumption. Therefore, in this work, we propose two approximate adder configurations that utilize novel logic optimization techniques to lower the power consumption and the hardware complexity of the circuit. The proposed approximate adders are designed using 55 nm technology and evaluated based on power consumption, delay, area, and power delay product (PDP). The simulation results indicate a reduction of 46.9% and 57.21% in power consumption for the approximate adder-1 & adder-2 compared to the conventional full adder. Furthermore, to validate the reliability of the proposed design, we also evaluated and calculated the accuracy metrics in terms of mean error distance (MED) of 0.25, which reflects the error tolerance of the proposed design.
Authors - Parambrata Sanyal, Mukund Kuthe, Sudhanshu Maurya, Sushmit Partakke, Firdous Sadaf M. Ismail, Rachit Garg Abstract - The most important public health challenge of myocardial infarction is caused by the obstruction by cholesterol and plaque accumulation in arteries, resulting in morbidity and mortality across the globe, especially in low and middle economies that lack health services, preventive measures, and early detection facilities. This study seeks to support the development of effective strategies by proposing a stacking ensemble model for timely forecasting and treatment of this disease in a serious way to improve healthcare significantly around the globe. The proposed methodology has been implemented on a retrospective dataset acquired from IEEE Dataport. The methodology involves normalization and standardization of the dataset, ensuring uniformity so that the machine learning classifiers work well. Our research compares several widely used machine learning classifiers, including Support Vector Machines (SVM), Gradient Boosting (GB), and Naive Bayes (NB), whose hyperparameter tuning has been done by grid search CV (GCV). The proposed stacking ensemble model stacks Light Gradient Boost and Cat Boost algorithms after being hyper-tuned by the Particle Swarm Optimization technique to enhance the overall predictive capacity. The results demonstrate that the proposed stacking ensemble model surpasses the individual classifiers in metrics, including the F1 score, recall, accuracy, and precision that are considered in this paper. Future directions of the research would be to work on expanded datasets and, most importantly, increase population diversity, add clinical parameters, and instead utilize more sophisticated machine learning techniques.
Authors - Mirza Zuber Baig, Vivek Nainwal, Anoop kumar, Bharat Kumar Abstract - Quantum computer simulators play a crucial role in understanding and analyzing the behavior of quantum systems. However, simulating large-scale quantum systems over classical machines can be computationally expensive and time consuming, limiting the practicality of many quantum algorithms. In this research paper, we explore the methodology employed for accelerating indigenous density-matrix based quantum computer simulator by using state of art libraries for GPUs (Graphics Processing Units) effectively increasing the number of Qubits it can simulate. The paper discusses the methods and techniques employed to identify computationally intensive and time-consuming functions within the simulator. By analyzing the profile results, we identified specific functions that required significant computational resources. To accelerate these functions, we utilized GPU acceleration techniques, leveraging parallel processing power. Our study demonstrates a significant improvement in simulation speed, achieving a significant speedup, showcasing the effectiveness of GPU acceleration in quantum computer simulations.
Authors - Ravi Kumar Suggala, Khushi Kumari, Mathi Gayathri, Koppisetti Deepika Naga Sree, Nekkalapudi Gayathri, Suma Kadali Abstract - Malware file production grows rather actively, which is explained by the development of digital structures. The proliferation of cyber trends poses severe security challenges due to the increasing complexity of attacks. These files could be difficult to detect when they share characteristics with normal files or if they are altered. Internet of Things (IoT) networks put a probability of vulnerability akin to Mirai malware to cyberattacks. There is a need to develop complex procedures for top security since it is important for such networks. This paper presents a new framework of preprocessing techniques, feature selection, and classification for predicting Mirai malware IoT security attacks. The preprocessing part uses the Global-Local Depth Normalization (GLDN) of features for dissolving noise and for better normalization of feature depths to enhance the learning factor. Practical feature selection is performed by using a combination of Gooseneck Barnacle Optimization (GBO) and Human Memory Optimization (HMO). This hybrid makes an intelligent dimensionality reduction decision determined by choosing appropriate features from among the set by the right balance between exploration and exploitation using biologically inspired optimization algorithms. For classification, there is proposed a Stereoscopic Scalable Quantum Convolutional Neural Network (sQCNN) that applies quantum computation principles to enhance computational scalability at the quantum level. The Banyan Tree Growth Optimization (BTGO) algorithm can optimize the classifier with high accuracy and attack detection immunity. The concept of Banyan tree growth in a hierarchical structure is similar to the classifier structure. Experiments conducted on the N-BaIoT dataset successfully prove the idea behind the proposed approach. The results propose that the new methods ensure better results over the traditional methods concerning the achieved accuracy of 99.67% and precision of 99.61%, while also incorporating reduced computational over- head. This new framework is a major step forward in defending IoT networks against current emerging threats, stressing the collaboration of preprocessing, feature selection, and quantum learning.
